IBL Specular complete.

CMakeLists.txt

@@ -9,9 +9,10 @@ set(CMAKE_CXX_STANDARD_REQUIRED ON)
 set(CMAKE_CXX_EXTENSIONS OFF)
 
 if (MSVC)
-    set(CMAKE_CXX_FLAGS "/W4 /GR- /Zi")
+    set(CMAKE_CXX_FLAGS "/W4 /GR- ${MSVC_FLAGS}")
     add_compile_definitions(_HAS_EXCEPTIONS=0)
     add_compile_definitions(_CRT_SECURE_NO_WARNINGS)
+    add_compile_definitions(${MSVC_DEFINES})
 else ()
     set(CMAKE_CXX_FLAGS "-Wall -g -fno-rtti -fno-exceptions")
 endif ()

CMakePresets.json

@@ -1,5 +1,5 @@
 {
-  "version": 2,
+  "version": 6,
   "configurePresets": [
     {
       "name": "linux",
@@ -11,16 +11,83 @@
         "CMAKE_C_COMPILER": "/usr/bin/clang",
         "CMAKE_CXX_COMPILER": "/usr/bin/clang++",
         "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/vcpkg/scripts/buildsystems/vcpkg.cmake"
+      },
+      "condition": {
+        "type": "equals",
+        "lhs": "${hostSystemName}",
+        "rhs": "Linux"
       }
     },
     {
-      "name": "windows",
+      "name": "windows-debug",
       "generator": "Ninja",
-      "binaryDir": "${sourceDir}/build",
+      "binaryDir": "${sourceDir}/build/debug/",
       "cacheVariables": {
+        "CMAKE_BUILD_TYPE": "Debug",
         "CMAKE_EXPORT_COMPILE_COMMANDS": true,
         "CMAKE_MAKE_PROGRAM": "ninja",
-        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/vcpkg/scripts/buildsystems/vcpkg.cmake"
+        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/vcpkg/scripts/buildsystems/vcpkg.cmake",
+        "DXC_SHADER_FLAGS": "-Zi",
+        "GLSLC_SHADER_FLAGS": "-g"
+      },
+      "condition": {
+        "type": "equals",
+        "lhs": "${hostSystemName}",
+        "rhs": "Windows"
+      }
+    },
+    {
+      "name": "windows-reldebug",
+      "generator": "Ninja",
+      "binaryDir": "${sourceDir}/build/reldebug/",
+      "cacheVariables": {
+        "CMAKE_BUILD_TYPE": "RelWithDebInfo",
+        "CMAKE_EXPORT_COMPILE_COMMANDS": true,
+        "CMAKE_MAKE_PROGRAM": "ninja",
+        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/vcpkg/scripts/buildsystems/vcpkg.cmake",
+        "DXC_SHADER_FLAGS": "-Zi",
+        "GLSLC_SHADER_FLAGS": "-g"
+      },
+      "condition": {
+        "type": "equals",
+        "lhs": "${hostSystemName}",
+        "rhs": "Windows"
+      }
+    },
+    {
+      "name": "windows-reldebug-nobreak",
+      "generator": "Ninja",
+      "binaryDir": "${sourceDir}/build/reldebug/",
+      "cacheVariables": {
+        "CMAKE_BUILD_TYPE": "RelWithDebInfo",
+        "CMAKE_EXPORT_COMPILE_COMMANDS": true,
+        "CMAKE_MAKE_PROGRAM": "ninja",
+        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/vcpkg/scripts/buildsystems/vcpkg.cmake",
+        "DXC_SHADER_FLAGS": "-Zi",
+        "GLSLC_SHADER_FLAGS": "-g",
+        "MSVC_DEFINES": "ASTER_NO_BREAK"
+      },
+      "condition": {
+        "type": "equals",
+        "lhs": "${hostSystemName}",
+        "rhs": "Windows"
+      }
+    },
+    {
+      "name": "windows-release",
+      "generator": "Ninja",
+      "binaryDir": "${sourceDir}/build/release/",
+      "cacheVariables": {
+        "CMAKE_BUILD_TYPE": "Release",
+        "CMAKE_EXPORT_COMPILE_COMMANDS": true,
+        "CMAKE_MAKE_PROGRAM": "ninja",
+        "CMAKE_TOOLCHAIN_FILE": "${sourceDir}/vcpkg/scripts/buildsystems/vcpkg.cmake",
+        "MSVC_DEFINES": "ASTER_NDEBUG"
+      },
+      "condition": {
+        "type": "equals",
+        "lhs": "${hostSystemName}",
+        "rhs": "Windows"
       }
     }
   ]

add_shader.cmake

@@ -16,7 +16,7 @@ function(add_shader TARGET SHADER)
         message("Marked as hlsl file. ${current-output-path}")
         add_custom_command(
                 OUTPUT ${current-output-path}
-                COMMAND Vulkan::dxc_exe -spirv -T "${shader-type}_6_0" -E main ${current-shader-path} -Fo ${current-output-path}
+                COMMAND Vulkan::dxc_exe ${DXC_SHADER_FLAGS} -spirv -T "${shader-type}_6_0" -E main ${current-shader-path} -Fo ${current-output-path}
                 DEPENDS ${current-shader-path}
                 IMPLICIT_DEPENDS CXX ${current-shader-path}
                 VERBATIM)
@@ -24,7 +24,7 @@ function(add_shader TARGET SHADER)
         message("Marked as glsl file. ${current-output-path}")
         add_custom_command(
                 OUTPUT ${current-output-path}
-                COMMAND Vulkan::glslc -o ${current-output-path} ${current-shader-path}
+                COMMAND Vulkan::glslc ${GLSLC_SHADER_FLAGS} -o ${current-output-path} ${current-shader-path}
                 DEPENDS ${current-shader-path}
                 IMPLICIT_DEPENDS CXX ${current-shader-path}
                 VERBATIM)

aster/config.h

@@ -20,7 +20,7 @@
 
 #define EASTL_NO_EXCEPTIONS 1
 
-#if defined(NDEBUG)
+#if defined(ASTER_NDEBUG)
 #define USE_OPTICK (0)
 #else
 #define USE_OPTICK (1)

aster/context.h

@@ -21,12 +21,18 @@ struct Context final
     vk::DebugUtilsMessengerEXT m_DebugMessenger = nullptr;
 
     // Ctor/Dtor
-    Context(cstr appName, Version version, bool enableValidation = true);
+    Context(cstr appName, Version version, bool enableValidation = ENABLE_VALIDATION_DEFAULT_VALUE);
     ~Context();
 
     // Move
     Context(Context &&other) noexcept;
     Context &operator=(Context &&other) noexcept;
 
+#if !defined(ASTER_NDEBUG)
+    constexpr static bool ENABLE_VALIDATION_DEFAULT_VALUE = true;
+#else
+    constexpr static bool ENABLE_VALIDATION_DEFAULT_VALUE = false;
+#endif
+
     DISALLOW_COPY_AND_ASSIGN(Context);
 };

aster/device.cpp

@@ -28,6 +28,7 @@ Device::Device(const Context *context, PhysicalDevice *physicalDevice, Features
                NameString &&name)
     : m_Name(std::move(name))
     , m_PhysicalDevice(physicalDevice->m_PhysicalDevice)
+    , m_ValidationEnabled(context->m_DebugMessenger != nullptr)
 {
     // Shouldn't have more than 4 deviceQueueFamilies in use anyway. Else we can heap
     eastl::fixed_vector<vk::DeviceQueueCreateInfo, 4> deviceQueueCreateInfos;

aster/device.h

@@ -29,6 +29,7 @@ struct Device final
     vk::Device m_Device = nullptr;
     VmaAllocator m_Allocator = nullptr;
     vk::PipelineCache m_PipelineCache = nullptr;
+    bool m_ValidationEnabled = true;
 
     template <typename T>
     requires vk::isVulkanHandleType<T>::value void SetName(const T &object, cstr name) const;
@@ -54,8 +55,9 @@ template <typename T>
 requires vk::isVulkanHandleType<T>::value void
 Device::SetName(const T &object, cstr name) const
 {
-    if (!name)
+    if (!m_ValidationEnabled || !name || !object)
         return;
+
     auto handle = Recast<u64>(Cast<typename T::NativeType>(object));
     const vk::DebugUtilsObjectNameInfoEXT objectNameInfo = {
         .objectType = object.objectType,

aster/global.h

@@ -23,7 +23,9 @@
 #undef min
 #endif
 
+#if !defined(NDEBUG)
 #define VULKAN_HPP_ASSERT(expr) DEBUG_IF(!(expr), "Vulkan assert failed")
+#endif 
 #include <EASTL/fixed_string.h>
 #include <EASTL/string.h>
 #include <vk_mem_alloc.h>
@@ -51,7 +53,7 @@ Failed(const vk::Result result)
 using NameString = eastl::fixed_string<char, 32, false>;
 
 template <typename TFlagBits>
-struct std::hash<vk::Flags<TFlagBits>> // NOLINT(*-dcl58-cpp)
+struct eastl::hash<vk::Flags<TFlagBits>> // NOLINT(*-dcl58-cpp)
 {
     [[nodiscard]] usize
     operator()(const vk::Flags<TFlagBits> &val)
@@ -64,7 +66,7 @@ template <typename T>
 [[nodiscard]] usize
 HashAny(const T &val)
 {
-    return std::hash<std::remove_cvref_t<T>>()(val);
+    return eastl::hash<std::remove_cvref_t<T>>()(val);
 }
 
 [[nodiscard]] inline usize

aster/image.cpp

@@ -11,7 +11,10 @@ void
 Image::Destroy(const Device *device)
 {
     if (!IsValid() || !IsOwned())
+    {
+        m_MipLevels_ = 0;
         return;
+    }
 
     device->m_Device.destroy(m_View, nullptr);
     vmaDestroyImage(device->m_Allocator, m_Image, m_Allocation);
@@ -280,7 +283,6 @@ void
 StorageTexture::Init(const Device *device, vk::Extent2D extent, const vk::Format imageFormat, const bool isSampled,
                      cstr name)
 {
-
     // Reasoning:
     // Transfer Src and Dst to copy to and from the buffer since Storage will often be loaded with info, and read for
     // results.
@@ -342,3 +344,72 @@ StorageTexture::Init(const Device *device, vk::Extent2D extent, const vk::Format
 
     device->SetName(m_Image, name);
 }
+
+void
+StorageTextureCube::Init(const Device* device, u32 cubeSide, vk::Format imageFormat, bool isSampled, bool isMipMapped, cstr name)
+{
+    // Reasoning:
+    // Transfer Src and Dst to copy to and from the buffer since Storage will often be loaded with info, and read for
+    // results.
+    auto usage =
+        vk::ImageUsageFlagBits::eStorage | vk::ImageUsageFlagBits::eTransferSrc | vk::ImageUsageFlagBits::eTransferDst;
+    if (isSampled)
+    {
+        WARN_IF(!IsPowerOfTwo(cubeSide), "Image {1} is {0}x{0} (Non Power of Two)",
+                cubeSide, name ? name : "<unnamed>");
+        usage |= vk::ImageUsageFlagBits::eSampled;
+    }
+
+    const u32 mipLevels = isMipMapped ? 1 + Cast<u32>(floor(log2(cubeSide))) : 1;
+    assert(mipLevels <= MIP_MASK);
+
+    vk::ImageCreateInfo imageCreateInfo = {
+        .flags = vk::ImageCreateFlagBits::eCubeCompatible,
+        .imageType = vk::ImageType::e2D,
+        .format = imageFormat,
+        .extent = {cubeSide, cubeSide, 1},
+        .mipLevels = mipLevels,
+        .arrayLayers = 6,
+        .samples = vk::SampleCountFlagBits::e1,
+        .tiling = vk::ImageTiling::eOptimal,
+        .usage = usage,
+        .sharingMode = vk::SharingMode::eExclusive,
+        .initialLayout = vk::ImageLayout::eUndefined,
+    };
+    constexpr VmaAllocationCreateInfo allocationCreateInfo = {
+        .flags = {},
+        .usage = VMA_MEMORY_USAGE_AUTO,
+    };
+
+    VkImage image;
+    VmaAllocation allocation;
+    auto result = Cast<vk::Result>(vmaCreateImage(device->m_Allocator, Recast<VkImageCreateInfo *>(&imageCreateInfo),
+                                                  &allocationCreateInfo, &image, &allocation, nullptr));
+    ERROR_IF(Failed(result), "Could not allocate image {}. Cause: {}", name, result) THEN_ABORT(result);
+
+    vk::ImageView view;
+    const vk::ImageViewCreateInfo imageViewCreateInfo = {
+        .image = image,
+        .viewType = vk::ImageViewType::eCube,
+        .format = imageFormat,
+        .components = {},
+        .subresourceRange =
+            {
+                .aspectMask = vk::ImageAspectFlagBits::eColor,
+                .baseMipLevel = 0,
+                .levelCount = mipLevels,
+                .baseArrayLayer = 0,
+                .layerCount = 6,
+            },
+    };
+    result = device->m_Device.createImageView(&imageViewCreateInfo, nullptr, &view);
+    ERROR_IF(Failed(result), "Could not create image view {}. Cause: {}", name, result) THEN_ABORT(result);
+
+    m_Image = image;
+    m_View = view;
+    m_Allocation = allocation;
+    m_Extent = imageCreateInfo.extent;
+    m_MipLevels_ = mipLevels | OWNED_BIT | VALID_BIT;
+
+    device->SetName(m_Image, name);
+}

aster/image.h

@@ -59,27 +59,45 @@ struct Texture : Image
     void Init(const Device *device, vk::Extent2D extent, vk::Format imageFormat, bool isMipMapped, cstr name = nullptr);
 };
 
+static_assert(sizeof(Texture) == sizeof(Image));
+
 struct TextureCube : Texture
 {
     void
     Init(const Device *device, u32 cubeSide, vk::Format imageFormat, bool isMipMapped = false, cstr name = nullptr);
 };
 
+static_assert(sizeof(TextureCube) == sizeof(Image));
+
 struct AttachmentImage : Image
 {
     void Init(const Device *device, vk::Extent2D extent, vk::Format imageFormat, cstr name = nullptr);
 };
 
+static_assert(sizeof(AttachmentImage) == sizeof(Image));
+
 struct DepthImage : Image
 {
     void Init(const Device *device, vk::Extent2D extent, cstr name = nullptr);
 };
 
+static_assert(sizeof(DepthImage) == sizeof(Image));
+
 struct StorageTexture : Texture
 {
     void Init(const Device *device, vk::Extent2D extent, vk::Format imageFormat, bool isSampled, cstr name = nullptr);
 };
 
+static_assert(sizeof(StorageTexture) == sizeof(Image));
+
+struct StorageTextureCube : StorageTexture
+{
+    void Init(const Device *device, u32 cubeSide, vk::Format imageFormat, bool isSampled, bool isMipMapped = false,
+              cstr name = nullptr);
+};
+
+static_assert(sizeof(StorageTextureCube) == sizeof(Image));
+
 inline bool
 Image::IsValid() const
 {

aster/logger.cpp

@@ -15,6 +15,5 @@ void
 AssertionFailure(const char *af)
 {
     ERROR("{}", af);
-    debug_break();
 }
 } // namespace eastl

aster/logger.h

@@ -69,12 +69,12 @@ struct Logger
             fmt::println("{}{} {} {} at {}:{}{}", ToColorCstr<TLogLevel>(), ToCstr<TLogLevel>(), message.data(),
                          ansi_color::Black, loc, line, ansi_color::Reset);
         }
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG) && !defined(ASTER_NO_BREAK)
         if constexpr (TLogLevel == LogType::eError)
         {
             debug_break();
         }
-#endif // !defined(NDEBUG)
+#endif
     }
 
     template <LogType TLogLevel>
@@ -86,12 +86,12 @@ struct Logger
             fmt::println("{}{} ({}) {} {} at {}:{}{}", ToColorCstr<TLogLevel>(), ToCstr<TLogLevel>(), exprStr,
                          message.data(), ansi_color::Black, loc, line, ansi_color::Reset);
         }
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG) && !defined(ASTER_NO_BREAK)
         if constexpr (TLogLevel == LogType::eError)
         {
             debug_break();
         }
-#endif // !defined(NDEBUG)
+#endif
     }
 };
 
@@ -135,7 +135,7 @@ extern Logger g_Logger;
     ;                                                                                                                  \
     else g_Logger.Log<Logger::LogType::eInfo>(fmt::format(__VA_ARGS__), __FILE__, __LINE__)
 
-#if !defined(DEBUG_LOG_DISABLED) && !defined(NDEBUG)
+#if !defined(DEBUG_LOG_DISABLED) && !defined(ASTER_NDEBUG)
 
 #define DEBUG(...) g_Logger.Log<Logger::LogType::eDebug>(fmt::format(__VA_ARGS__), __FILE__, __LINE__)
 #define DEBUG_IF(expr, ...)                                                                                            \
@@ -151,24 +151,26 @@ extern Logger g_Logger;
 
 #else // !defined(DEBUG_LOG_DISABLED)
 
-#define DEBUG(msg)                                                                                                     \
+#define DEBUG(...)                                                                                                     \
     {                                                                                                                  \
     }
-#define DEBUG_IF(expr, msg)                                                                                            \
-    if (expr)                                                                                                          \
-    (void)msg
-#define ELSE_IF_DEBUG(expr, msg)                                                                                       \
+#define DEBUG_IF(expr, ...)                                                                                            \
+    if (false)                                                                                                         \
+    {                                                                                                                  \
+    }
+#define ELSE_IF_DEBUG(expr, ...)                                                                                       \
     ;                                                                                                                  \
-    if (expr)                                                                                                          \
-    (void)msg
-#define ELSE_DEBUG(msg)                                                                                                \
+    if (false)                                                                                                         \
+    {                                                                                                                  \
+    }
+#define ELSE_DEBUG(...)                                                                                                \
     ;                                                                                                                  \
     {                                                                                                                  \
     }
 
 #endif // !defined(DEBUG_LOG_DISABLED)
 
-#if !defined(VERBOSE_LOG_DISABLED) && !defined(NDEBUG)
+#if !defined(VERBOSE_LOG_DISABLED) && !defined(ASTER_NDEBUG)
 
 #define VERBOSE(...) g_Logger.Log<Logger::LogType::eVerbose>(fmt::format(__VA_ARGS__), __FILE__, __LINE__)
 #define VERBOSE_IF(expr, ...)                                                                                          \
@@ -184,17 +186,20 @@ extern Logger g_Logger;
 
 #else // !defined(DEBUG_LOG_DISABLED)
 
-#define VERBOSE(msg)                                                                                                   \
+#define VERBOSE(...)                                                                                                   \
     {                                                                                                                  \
     }
-#define VERBOSE_IF(expr, msg)                                                                                          \
-    if (expr)                                                                                                          \
-    (void)msg
-#define ELSE_IF_VERBOSE(expr, msg)                                                                                     \
+#define VERBOSE_IF(expr, ...)                                                                                          \
+    if (false)                                                                                                         \
+    {                                                                                                                  \
+    }                                                                                                                  \
+    }
+#define ELSE_IF_VERBOSE(expr, ...)                                                                                     \
     ;                                                                                                                  \
-    if (expr)                                                                                                          \
-    (void)msg
-#define ELSE_VERBOSE(msg)                                                                                              \
+    if (false)                                                                                                         \
+    {                                                                                                                  \
+    }
+#define ELSE_VERBOSE(...)                                                                                              \
     ;                                                                                                                  \
     {                                                                                                                  \
     }

samples/00_util/gpu_resource_manager.cpp

@@ -22,7 +22,7 @@ TextureManager::Init(const u32 maxCapacity)
 TextureHandle
 TextureManager::Commit(Texture *texture)
 {
-    ERROR_IF(!texture || !texture->IsValid() || !texture->IsOwned(), "Texture must be valid and owned for commital")
+    ERROR_IF(!texture || !texture->IsValid(), "Texture must be valid for commital")
     THEN_ABORT(-1);
 
     if (m_FreeHead != GpuResourceHandle::INVALID_HANDLE)
@@ -193,6 +193,79 @@ StorageTextureManager::Release(const Device *device, const StorageTextureHandle
     TextureManager::Release(device, {handle.m_Index});
 }
 
+usize
+HashSamplerCreateInfo(const vk::SamplerCreateInfo *createInfo)
+{
+    usize hash = HashAny(createInfo->flags);
+    hash = HashCombine(hash, HashAny(createInfo->magFilter));
+    hash = HashCombine(hash, HashAny(createInfo->minFilter));
+    hash = HashCombine(hash, HashAny(createInfo->mipmapMode));
+    hash = HashCombine(hash, HashAny(createInfo->addressModeU));
+    hash = HashCombine(hash, HashAny(createInfo->addressModeV));
+    hash = HashCombine(hash, HashAny(createInfo->addressModeW));
+    hash = HashCombine(hash, HashAny(Cast<usize>(createInfo->mipLodBias * 1000))); // Resolution of 10^-3
+    hash = HashCombine(hash, HashAny(createInfo->anisotropyEnable));
+    hash = HashCombine(hash,
+                       HashAny(Cast<usize>(createInfo->maxAnisotropy * 0x10))); // 16:1 Anisotropy is enough resolution
+    hash = HashCombine(hash, HashAny(createInfo->compareEnable));
+    hash = HashCombine(hash, HashAny(createInfo->compareOp));
+    hash = HashCombine(hash, HashAny(Cast<usize>(createInfo->minLod * 1000))); // 0.001 resolution is enough.
+    hash = HashCombine(hash,
+                       HashAny(Cast<usize>(createInfo->maxLod * 1000))); // 0.001 resolution is enough. (1 == NO Clamp)
+    hash = HashCombine(hash, HashAny(createInfo->borderColor));
+    hash = HashCombine(hash, HashAny(createInfo->unnormalizedCoordinates));
+
+    return hash;
+}
+
+void
+SamplerManager::Init(usize size)
+{
+    m_Samplers.reserve(size);
+    m_SamplerHashes.reserve(size);
+}
+
+SamplerHandle
+SamplerManager::Create(const Device *device, const vk::SamplerCreateInfo *createInfo)
+{
+    const usize hash = HashSamplerCreateInfo(createInfo);
+
+    for (u32 index = 0; usize samplerHash : m_SamplerHashes)
+    {
+        if (samplerHash == hash)
+        {
+            return {index};
+        }
+        ++index;
+    }
+
+    vk::Sampler sampler;
+    AbortIfFailed(device->m_Device.createSampler(createInfo, nullptr, &sampler));
+    const u32 index = Cast<u32>(m_SamplerHashes.size());
+    m_SamplerHashes.push_back(hash);
+    m_Samplers.push_back(sampler);
+    return {index};
+}
+
+vk::Sampler
+SamplerManager::Fetch(const SamplerHandle handle)
+{
+    assert(!handle.IsInvalid());
+
+    return m_Samplers[handle.m_Index];
+}
+
+void
+SamplerManager::Destroy(const Device *device)
+{
+    for (const auto &sampler : m_Samplers)
+    {
+        device->m_Device.destroy(sampler, nullptr);
+    }
+    m_Samplers.clear();
+    m_SamplerHashes.clear();
+}
+
 GpuResourceManager::WriteInfo::WriteInfo(vk::DescriptorBufferInfo info)
     : uBufferInfo(info)
 {
@@ -230,7 +303,7 @@ GpuResourceManager::Commit(StorageBuffer *storageBuffer)
 
     m_WriteOwner.emplace_back(HandleType::eBuffer, handle.m_Index);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     ++m_CommitedBufferCount;
 #endif
 
@@ -276,7 +349,7 @@ GpuResourceManager::Release(BufferHandle handle)
 
     m_BufferManager.Release(m_Device, handle);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     --m_CommitedBufferCount;
 #endif
 }
@@ -304,7 +377,7 @@ GpuResourceManager::Release(TextureHandle handle)
 
     m_TextureManager.Release(m_Device, handle);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     --m_CommitedTextureCount;
 #endif
 }
@@ -323,12 +396,14 @@ GpuResourceManager::Release(Texture *texture, TextureHandle handle)
 }
 
 TextureHandle
-GpuResourceManager::Commit(Texture *texture)
+GpuResourceManager::CommitTexture(Texture *texture, const SamplerHandle sampler)
 {
     TextureHandle handle = m_TextureManager.Commit(texture);
 
+    const vk::Sampler samplerImpl = sampler.IsInvalid() ? m_DefaultSampler : m_SamplerManager.Fetch(sampler);
+
     m_WriteInfos.emplace_back(vk::DescriptorImageInfo{
-        .sampler = nullptr,
+        .sampler = samplerImpl,
         .imageView = texture->m_View,
         .imageLayout = vk::ImageLayout::eShaderReadOnlyOptimal,
     });
@@ -344,7 +419,7 @@ GpuResourceManager::Commit(Texture *texture)
 
     m_WriteOwner.emplace_back(HandleType::eTexture, handle.m_Index);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     ++m_CommitedTextureCount;
 #endif
 
@@ -352,12 +427,14 @@ GpuResourceManager::Commit(Texture *texture)
 }
 
 StorageTextureHandle
-GpuResourceManager::Commit(StorageTexture *storageTexture)
+GpuResourceManager::CommitStorageTexture(StorageTexture *storageTexture, SamplerHandle sampler)
 {
     StorageTextureHandle handle = m_StorageTextureManager.Commit(storageTexture);
 
+    vk::Sampler samplerImpl = sampler.IsInvalid() ? m_DefaultSampler : m_SamplerManager.Fetch(sampler);
+
     m_WriteInfos.emplace_back(vk::DescriptorImageInfo{
-        .sampler = nullptr,
+        .sampler = samplerImpl,
         .imageView = storageTexture->m_View,
         .imageLayout = vk::ImageLayout::eGeneral,
     });
@@ -373,7 +450,7 @@ GpuResourceManager::Commit(StorageTexture *storageTexture)
 
     m_WriteOwner.emplace_back(HandleType::eStorageTexture, handle.m_Index);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     ++m_CommitedStorageTextureCount;
 #endif
 
@@ -390,7 +467,7 @@ GpuResourceManager::Release(StorageTextureHandle handle)
 
     m_StorageTextureManager.Release(m_Device, handle);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     --m_CommitedStorageTextureCount;
 #endif
 }
@@ -432,8 +509,16 @@ GpuResourceManager::GpuResourceManager(Device *device, u16 maxSize)
     u32 storageTexturesCount =
         eastl::min(properties.limits.maxPerStageDescriptorStorageImages - 1024, Cast<u32>(maxSize));
 
-    // TODO: Switch to bindless samplers / multiple sampler configurations
-    vk::SamplerCreateInfo samplerCreateInfo = {
+    INFO("Max Buffer Count: {}", buffersCount);
+    INFO("Max Texture Count: {}", texturesCount);
+    INFO("Max Storage Texture Count: {}", storageTexturesCount);
+
+    m_BufferManager.Init(buffersCount);
+    m_TextureManager.Init(texturesCount);
+    m_StorageTextureManager.Init(storageTexturesCount);
+    m_SamplerManager.Init(storageTexturesCount);
+
+    m_DefaultSamplerCreateInfo = {
         .magFilter = vk::Filter::eLinear,
         .minFilter = vk::Filter::eLinear,
         .mipmapMode = vk::SamplerMipmapMode::eLinear,
@@ -449,15 +534,8 @@ GpuResourceManager::GpuResourceManager(Device *device, u16 maxSize)
         .borderColor = vk::BorderColor::eFloatOpaqueBlack,
         .unnormalizedCoordinates = false,
     };
-    AbortIfFailed(device->m_Device.createSampler(&samplerCreateInfo, nullptr, &m_ImmutableSampler));
 
-    INFO("Max Buffer Count: {}", buffersCount);
-    INFO("Max Texture Count: {}", texturesCount);
-    INFO("Max Storage Texture Count: {}", storageTexturesCount);
-
-    m_BufferManager.Init(buffersCount);
-    m_TextureManager.Init(texturesCount);
-    m_StorageTextureManager.Init(storageTexturesCount);
+    m_DefaultSampler = m_SamplerManager.Fetch(m_SamplerManager.Create(device, &m_DefaultSamplerCreateInfo));
 
     eastl::array poolSizes = {
         vk::DescriptorPoolSize{
@@ -495,7 +573,6 @@ GpuResourceManager::GpuResourceManager(Device *device, u16 maxSize)
         .pBindingFlags = layoutBindingFlags.data(),
     };
 
-    eastl::vector immutableSamplers(texturesCount, m_ImmutableSampler);
     eastl::array descriptorLayoutBindings = {
         vk::DescriptorSetLayoutBinding{
             .binding = BUFFER_BINDING_INDEX,
@@ -508,7 +585,6 @@ GpuResourceManager::GpuResourceManager(Device *device, u16 maxSize)
             .descriptorType = vk::DescriptorType::eCombinedImageSampler,
             .descriptorCount = Cast<u32>(texturesCount),
             .stageFlags = vk::ShaderStageFlagBits::eAll,
-            .pImmutableSamplers = immutableSamplers.data(),
         },
         vk::DescriptorSetLayoutBinding{
             .binding = STORAGE_TEXTURE_BINDING_INDEX,
@@ -543,7 +619,7 @@ GpuResourceManager::GpuResourceManager(Device *device, u16 maxSize)
 
 GpuResourceManager::~GpuResourceManager()
 {
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     WARN_IF(m_CommitedBufferCount > 0 || m_CommitedTextureCount > 0 || m_CommitedStorageTextureCount > 0,
             "Resources alive: SSBO = {}, Textures = {}, RWTexture = {}", m_CommitedBufferCount, m_CommitedTextureCount,
             m_CommitedStorageTextureCount);
@@ -552,7 +628,7 @@ GpuResourceManager::~GpuResourceManager()
     m_BufferManager.Destroy(m_Device);
     m_TextureManager.Destroy(m_Device);
     m_StorageTextureManager.Destroy(m_Device);
-    m_Device->m_Device.destroy(m_ImmutableSampler, nullptr);
+    m_SamplerManager.Destroy(m_Device);
     m_Device->m_Device.destroy(m_DescriptorPool, nullptr);
     m_Device->m_Device.destroy(m_SetLayout, nullptr);
 }
@@ -561,16 +637,19 @@ GpuResourceManager::GpuResourceManager(GpuResourceManager &&other) noexcept
     : m_WriteInfos(std::move(other.m_WriteInfos))
     , m_Writes(std::move(other.m_Writes))
     , m_WriteOwner(std::move(other.m_WriteOwner))
-    , m_ImmutableSampler(other.m_ImmutableSampler)
     , m_BufferManager(std::move(other.m_BufferManager))
     , m_TextureManager(std::move(other.m_TextureManager))
     , m_StorageTextureManager(std::move(other.m_StorageTextureManager))
+    , m_SamplerManager(std::move(other.m_SamplerManager))
     , m_Device(Take(other.m_Device))
     , m_DescriptorPool(other.m_DescriptorPool)
     , m_SetLayout(other.m_SetLayout)
     , m_DescriptorSet(other.m_DescriptorSet)
+#if !defined(ASTER_NDEBUG)
     , m_CommitedBufferCount(other.m_CommitedBufferCount)
     , m_CommitedTextureCount(other.m_CommitedTextureCount)
+    , m_CommitedStorageTextureCount(other.m_CommitedStorageTextureCount)
+#endif
 {
     assert(!other.m_Device);
 }
@@ -583,17 +662,26 @@ GpuResourceManager::operator=(GpuResourceManager &&other) noexcept
     m_WriteInfos = std::move(other.m_WriteInfos);
     m_Writes = std::move(other.m_Writes);
     m_WriteOwner = std::move(other.m_WriteOwner);
-    m_ImmutableSampler = other.m_ImmutableSampler;
     m_BufferManager = std::move(other.m_BufferManager);
     m_TextureManager = std::move(other.m_TextureManager);
     m_StorageTextureManager = std::move(other.m_StorageTextureManager);
+    m_SamplerManager = std::move(other.m_SamplerManager);
     m_Device = Take(other.m_Device); // Ensure taken.
     m_DescriptorPool = other.m_DescriptorPool;
     m_SetLayout = other.m_SetLayout;
     m_DescriptorSet = other.m_DescriptorSet;
+#if !defined(ASTER_NDEBUG)
     m_CommitedBufferCount = other.m_CommitedBufferCount;
     m_CommitedTextureCount = other.m_CommitedTextureCount;
+    m_CommitedStorageTextureCount = other.m_CommitedStorageTextureCount;
+#endif
 
     assert(!other.m_Device);
     return *this;
 }
+
+SamplerHandle
+GpuResourceManager::CreateSampler(const vk::SamplerCreateInfo *samplerCreateInfo)
+{
+    return m_SamplerManager.Create(m_Device, samplerCreateInfo);
+}

samples/00_util/gpu_resource_manager.h

@@ -39,6 +39,10 @@ struct StorageTextureHandle : GpuResourceHandle
 {
 };
 
+struct SamplerHandle : GpuResourceHandle
+{
+};
+
 struct TextureManager
 {
     eastl::vector<Texture> m_Textures;
@@ -72,6 +76,18 @@ struct StorageTextureManager : TextureManager
     void Release(const Device *device, StorageTextureHandle handle);
 };
 
+struct SamplerManager
+{
+    // There can only be so many samplers.
+    eastl::vector<vk::Sampler> m_Samplers;
+    eastl::vector<usize> m_SamplerHashes;
+
+    void Init(usize size);
+    SamplerHandle Create(const Device *device, const vk::SamplerCreateInfo *createInfo);
+    vk::Sampler Fetch(SamplerHandle handle);
+    void Destroy(const Device *device);
+};
+
 struct GpuResourceManager
 {
   private:
@@ -102,11 +118,12 @@ struct GpuResourceManager
     eastl::vector<vk::WriteDescriptorSet> m_Writes;
     eastl::vector<WriteOwner> m_WriteOwner;
 
-    vk::Sampler m_ImmutableSampler;
+    vk::Sampler m_DefaultSampler;
 
     BufferManager m_BufferManager;
     TextureManager m_TextureManager;
     StorageTextureManager m_StorageTextureManager;
+    SamplerManager m_SamplerManager;
 
     void EraseWrites(u32 handleIndex, HandleType handleType);
 
@@ -117,6 +134,8 @@ struct GpuResourceManager
     constexpr static u32 TEXTURE_BINDING_INDEX = 1;
     constexpr static u32 STORAGE_TEXTURE_BINDING_INDEX = 2;
 
+    vk::SamplerCreateInfo m_DefaultSamplerCreateInfo;
+
     vk::DescriptorPool m_DescriptorPool;
     vk::DescriptorSetLayout m_SetLayout;
     vk::DescriptorSet m_DescriptorSet;
@@ -126,13 +145,16 @@ struct GpuResourceManager
     void Release(BufferHandle handle);                                           // Release and Destroy
     void Release(StorageBuffer *storageBuffer, BufferHandle handle);             // Release and Return
 
-    TextureHandle Commit(Texture *texture);               // Commit to GPU and take Ownership
-    void Release(TextureHandle handle);                   // Release and Destroy
-    void Release(Texture *texture, TextureHandle handle); // Release and Return
+    TextureHandle CommitTexture(Texture *texture, SamplerHandle sampler = {}); // Commit to GPU and take Ownership
+    void Release(TextureHandle handle);                                        // Release and Destroy
+    void Release(Texture *texture, TextureHandle handle);                      // Release and Return
 
-    StorageTextureHandle Commit(StorageTexture *storageTexture);
-    void Release(StorageTextureHandle handle);
-    void Release(StorageTexture *texture, StorageTextureHandle handle);
+    StorageTextureHandle
+    CommitStorageTexture(StorageTexture *storageTexture, SamplerHandle sampler = {}); // Commit to GPU and take Ownership
+    void Release(StorageTextureHandle handle);                                   // Release and Destroy
+    void Release(StorageTexture *texture, StorageTextureHandle handle);          // Release and Return
+
+    SamplerHandle CreateSampler(const vk::SamplerCreateInfo *samplerCreateInfo);
 
     void Update(); // Update all the descriptors required.
 
@@ -143,7 +165,7 @@ struct GpuResourceManager
     GpuResourceManager(GpuResourceManager &&other) noexcept;
     GpuResourceManager &operator=(GpuResourceManager &&other) noexcept;
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     usize m_CommitedBufferCount = 0;
     usize m_CommitedTextureCount = 0;
     usize m_CommitedStorageTextureCount = 0;

samples/00_util/gui.cpp

@@ -165,7 +165,7 @@ Draw(const vk::CommandBuffer commandBuffer, const vk::Extent2D extent, const vk:
 {
     // OPTICK_EVENT();
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     constexpr vk::DebugUtilsLabelEXT label = {
         .pLabelName = "UI pass",
         .color = std::array{0.9f, 0.9f, 1.0f, 1.0f},
@@ -195,7 +195,7 @@ Draw(const vk::CommandBuffer commandBuffer, const vk::Extent2D extent, const vk:
 
     commandBuffer.endRendering();
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     commandBuffer.endDebugUtilsLabelEXT();
 #endif
 }

samples/03_model_render/CMakeLists.txt

@@ -19,9 +19,12 @@ add_executable(model_render model_render.cpp
 
 add_shader(model_render shader/model.vs.hlsl)
 add_shader(model_render shader/model.ps.hlsl)
-add_shader(model_render shader/eqrectToCube.cs.hlsl)
+add_shader(model_render shader/eqrect_to_cube.cs.hlsl)
 add_shader(model_render shader/background.vs.hlsl)
 add_shader(model_render shader/background.ps.hlsl)
+add_shader(model_render shader/diffuse_irradiance.cs.hlsl)
+add_shader(model_render shader/prefilter.cs.hlsl)
+add_shader(model_render shader/brdf_lut.cs.hlsl)
 
 target_link_libraries(model_render PRIVATE aster_core)
 target_link_libraries(model_render PRIVATE util_helper)

samples/03_model_render/asset_loader.cpp

@@ -12,13 +12,13 @@
 
 #include "asset_loader.h"
 
+#include "EASTL/fixed_vector.h"
 #include "buffer.h"
 #include "device.h"
 #include "gpu_resource_manager.h"
 #include "helpers.h"
 #include "image.h"
 
-#include <EASTL/array.h>
 #include <EASTL/hash_map.h>
 #include <glm/gtc/type_ptr.hpp>
 
@@ -151,7 +151,7 @@ AssetLoader::LoadHdrImage(Texture *texture, cstr path, cstr name) const
 
     AbortIfFailed(m_CommandBuffer.begin(&OneTimeCmdBeginInfo));
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     StackString<128> loadActionName = "Load: ";
     loadActionName += name ? name : path;
     vk::DebugUtilsLabelEXT debugLabel = {
@@ -165,7 +165,7 @@ AssetLoader::LoadHdrImage(Texture *texture, cstr path, cstr name) const
     m_CommandBuffer.copyBufferToImage2(&stagingInfo);
     m_CommandBuffer.pipelineBarrier2(&postStagingDependency);
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     m_CommandBuffer.endDebugUtilsLabelEXT();
 #endif
 
@@ -190,6 +190,176 @@ AssetLoader::LoadHdrImage(Texture *texture, cstr path, cstr name) const
     stagingBuffer.Destroy(pDevice);
 }
 
+void
+GenerateMipMaps(vk::CommandBuffer commandBuffer, Texture *texture, vk::ImageLayout initialLayout,
+                vk::ImageLayout finalLayout, vk::PipelineStageFlags2 prevStage, vk::PipelineStageFlags2 finalStage)
+{
+#if !defined(ASTER_NDEBUG)
+    vk::DebugUtilsLabelEXT label = {
+        .pLabelName = "Generate Mipmap",
+        .color = std::array{0.9f, 0.9f, 0.9f, 1.0f},
+    };
+    commandBuffer.beginDebugUtilsLabelEXT(&label);
+#endif
+
+    vk::ImageMemoryBarrier2 imageStartBarrier = {
+        .srcStageMask = prevStage,
+        .srcAccessMask = vk::AccessFlagBits2::eNone,
+        .dstStageMask = vk::PipelineStageFlagBits2::eTransfer,
+        .dstAccessMask = vk::AccessFlagBits2::eTransferRead,
+        .oldLayout = initialLayout,
+        .newLayout = vk::ImageLayout::eTransferSrcOptimal,
+        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .image = texture->m_Image,
+        .subresourceRange =
+            {
+                .aspectMask = vk::ImageAspectFlagBits::eColor,
+                .baseMipLevel = 0,
+                .levelCount = 1,
+                .baseArrayLayer = 0,
+                .layerCount = 1,
+            },
+    };
+    vk::ImageMemoryBarrier2 mipsStartBarrier = imageStartBarrier;
+    mipsStartBarrier.dstAccessMask = vk::AccessFlagBits2::eTransferWrite;
+    mipsStartBarrier.oldLayout = vk::ImageLayout::eUndefined;
+    mipsStartBarrier.newLayout = vk::ImageLayout::eTransferDstOptimal;
+    mipsStartBarrier.subresourceRange = {
+        .aspectMask = vk::ImageAspectFlagBits::eColor,
+        .baseMipLevel = 1,
+        .levelCount = texture->GetMipLevels() - 1,
+        .baseArrayLayer = 0,
+        .layerCount = 1,
+    };
+    eastl::fixed_vector<vk::ImageMemoryBarrier2, 2> startBarriers = {
+        mipsStartBarrier,
+    };
+    if (initialLayout != imageStartBarrier.newLayout)
+    {
+        startBarriers.push_back(imageStartBarrier);
+    }
+
+    vk::DependencyInfo imageStartDependency = {
+        .imageMemoryBarrierCount = Cast<u32>(startBarriers.size()),
+        .pImageMemoryBarriers = startBarriers.data(),
+    };
+
+    vk::ImageMemoryBarrier2 nextMipBarrier = {
+        .srcStageMask = vk::PipelineStageFlagBits2::eTransfer,
+        .srcAccessMask = vk::AccessFlagBits2::eTransferWrite,
+        .dstStageMask = vk::PipelineStageFlagBits2::eTransfer,
+        .dstAccessMask = vk::AccessFlagBits2::eTransferRead,
+        .oldLayout = vk::ImageLayout::eTransferDstOptimal,
+        .newLayout = vk::ImageLayout::eTransferSrcOptimal,
+        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .image = texture->m_Image,
+        .subresourceRange =
+            {
+                .aspectMask = vk::ImageAspectFlagBits::eColor,
+                .baseMipLevel = 0,
+                .levelCount = 1,
+                .baseArrayLayer = 0,
+                .layerCount = 1,
+            },
+    };
+    vk::DependencyInfo interMipDependency = {
+        .imageMemoryBarrierCount = 1,
+        .pImageMemoryBarriers = &nextMipBarrier,
+    };
+
+    vk::ImageMemoryBarrier2 imageReadyBarrier = {
+        .srcStageMask = vk::PipelineStageFlagBits2::eTransfer,
+        .srcAccessMask = vk::AccessFlagBits2::eTransferWrite,
+        .dstStageMask = finalStage,
+        .dstAccessMask = vk::AccessFlagBits2::eShaderRead,
+        .oldLayout = vk::ImageLayout::eTransferSrcOptimal,
+        .newLayout = finalLayout,
+        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .image = texture->m_Image,
+        .subresourceRange =
+            {
+                .aspectMask = vk::ImageAspectFlagBits::eColor,
+                .baseMipLevel = 0,
+                .levelCount = texture->GetMipLevels(),
+                .baseArrayLayer = 0,
+                .layerCount = 1,
+            },
+    };
+    vk::DependencyInfo imageReadyDependency = {
+        .imageMemoryBarrierCount = 1,
+        .pImageMemoryBarriers = &imageReadyBarrier,
+    };
+
+    vk::ImageBlit2 blitRegion = {
+        .srcSubresource =
+            {
+                .aspectMask = vk::ImageAspectFlagBits::eColor,
+                .baseArrayLayer = 0,
+                .layerCount = 1,
+            },
+        .dstSubresource =
+            {
+                .aspectMask = vk::ImageAspectFlagBits::eColor,
+                .baseArrayLayer = 0,
+                .layerCount = 1,
+            },
+    };
+
+    vk::BlitImageInfo2 mipBlitInfo = {
+        .srcImage = texture->m_Image,
+        .srcImageLayout = vk::ImageLayout::eTransferSrcOptimal,
+        .dstImage = texture->m_Image,
+        .dstImageLayout = vk::ImageLayout::eTransferDstOptimal,
+        .regionCount = 1,
+        .pRegions = &blitRegion,
+        .filter = vk::Filter::eLinear,
+    };
+
+    auto calcNextMip = [](i32 prev) { return eastl::max(prev / 2, 1); };
+
+    // Mip Mapping
+
+    commandBuffer.pipelineBarrier2(&imageStartDependency);
+
+    i32 prevMipWidth = Cast<i32>(texture->m_Extent.width);
+    i32 prevMipHeight = Cast<i32>(texture->m_Extent.height);
+
+    u32 maxPrevMip = texture->GetMipLevels() - 1;
+    for (u32 prevMipLevel = 0; prevMipLevel < maxPrevMip; ++prevMipLevel)
+    {
+        i32 currentMipWidth = calcNextMip(prevMipWidth);
+        i32 currentMipHeight = calcNextMip(prevMipHeight);
+        u32 currentMipLevel = prevMipLevel + 1;
+
+        blitRegion.srcSubresource.mipLevel = prevMipLevel;
+        blitRegion.srcOffsets = std::array{
+            vk::Offset3D{0, 0, 0},
+            vk::Offset3D{prevMipWidth, prevMipHeight, 1},
+        };
+        blitRegion.dstSubresource.mipLevel = currentMipLevel;
+        blitRegion.dstOffsets = std::array{
+            vk::Offset3D{0, 0, 0},
+            vk::Offset3D{currentMipWidth, currentMipHeight, 1},
+        };
+
+        nextMipBarrier.subresourceRange.baseMipLevel = currentMipLevel;
+
+        commandBuffer.blitImage2(&mipBlitInfo);
+        commandBuffer.pipelineBarrier2(&interMipDependency);
+
+        prevMipHeight = currentMipHeight;
+        prevMipWidth = currentMipWidth;
+    }
+
+    commandBuffer.pipelineBarrier2(&imageReadyDependency);
+#if !defined(ASTER_NDEBUG)
+    commandBuffer.endDebugUtilsLabelEXT();
+#endif
+}
+
 TextureHandle
 AssetLoader::LoadImageToGpu(StagingBuffer *stagingBuffer, tinygltf::Image *image, bool isSrgb) const
 {
@@ -209,7 +379,7 @@ AssetLoader::LoadImageToGpu(StagingBuffer *stagingBuffer, tinygltf::Image *image
     stagingBuffer->Init(m_ResourceManager->m_Device, byteSize);
     stagingBuffer->Write(m_ResourceManager->m_Device, 0, byteSize, image->image.data());
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     StackString<128> loadActionName = "Load: ";
     loadActionName += image->name.empty() ? "<texture>" : image->name.c_str();
     vk::DebugUtilsLabelEXT debugLabel = {
@@ -235,7 +405,7 @@ AssetLoader::LoadImageToGpu(StagingBuffer *stagingBuffer, tinygltf::Image *image
             {
                 .aspectMask = vk::ImageAspectFlagBits::eColor,
                 .baseMipLevel = 0,
-                .levelCount = texture.GetMipLevels(),
+                .levelCount = 1,
                 .baseArrayLayer = 0,
                 .layerCount = 1,
             },
@@ -247,15 +417,13 @@ AssetLoader::LoadImageToGpu(StagingBuffer *stagingBuffer, tinygltf::Image *image
         .pImageMemoryBarriers = &imageStartBarrier,
     };
 
-    vk::ImageMemoryBarrier2 nextMipBarrier = {
-        .srcStageMask = vk::PipelineStageFlagBits2::eTransfer,
-        .srcAccessMask = vk::AccessFlagBits2::eTransferWrite,
-        .dstStageMask = vk::PipelineStageFlagBits2::eTransfer,
-        .dstAccessMask = vk::AccessFlagBits2::eTransferRead,
+    vk::ImageMemoryBarrier2 postStagingBarrier = {
+        .srcStageMask = vk::PipelineStageFlagBits2::eAllTransfer,
+        .dstStageMask = vk::PipelineStageFlagBits2::eAllTransfer,
         .oldLayout = vk::ImageLayout::eTransferDstOptimal,
         .newLayout = vk::ImageLayout::eTransferSrcOptimal,
-        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
+        .srcQueueFamilyIndex = m_TransferQueueIndex,
+        .dstQueueFamilyIndex = m_GraphicsQueueIndex,
         .image = texture.m_Image,
         .subresourceRange =
             {
@@ -266,33 +434,11 @@ AssetLoader::LoadImageToGpu(StagingBuffer *stagingBuffer, tinygltf::Image *image
                 .layerCount = 1,
             },
     };
-    vk::DependencyInfo interMipDependency = {
-        .imageMemoryBarrierCount = 1,
-        .pImageMemoryBarriers = &nextMipBarrier,
-    };
+    ;
 
-    vk::ImageMemoryBarrier2 imageReadyBarrier = {
-        .srcStageMask = vk::PipelineStageFlagBits2::eTransfer,
-        .srcAccessMask = vk::AccessFlagBits2::eTransferRead,
-        .dstStageMask = vk::PipelineStageFlagBits2::eFragmentShader,
-        .dstAccessMask = vk::AccessFlagBits2::eShaderRead,
-        .oldLayout = vk::ImageLayout::eTransferSrcOptimal,
-        .newLayout = vk::ImageLayout::eShaderReadOnlyOptimal,
-        .srcQueueFamilyIndex = m_TransferQueueIndex,
-        .dstQueueFamilyIndex = m_GraphicsQueueIndex,
-        .image = texture.m_Image,
-        .subresourceRange =
-            {
-                .aspectMask = vk::ImageAspectFlagBits::eColor,
-                .baseMipLevel = 0,
-                .levelCount = texture.GetMipLevels(),
-                .baseArrayLayer = 0,
-                .layerCount = 1,
-            },
-    };
-    vk::DependencyInfo imageReadyDependency = {
+    vk::DependencyInfo postStagingDependency = {
         .imageMemoryBarrierCount = 1,
-        .pImageMemoryBarriers = &imageReadyBarrier,
+        .pImageMemoryBarriers = &postStagingBarrier,
     };
 
     vk::BufferImageCopy2 imageCopy = {
@@ -317,78 +463,20 @@ AssetLoader::LoadImageToGpu(StagingBuffer *stagingBuffer, tinygltf::Image *image
         .pRegions = &imageCopy,
     };
 
-    vk::ImageBlit2 blitRegion = {
-        .srcSubresource =
-            {
-                .aspectMask = vk::ImageAspectFlagBits::eColor,
-                .baseArrayLayer = 0,
-                .layerCount = 1,
-            },
-        .dstSubresource =
-            {
-                .aspectMask = vk::ImageAspectFlagBits::eColor,
-                .baseArrayLayer = 0,
-                .layerCount = 1,
-            },
-    };
-
-    vk::BlitImageInfo2 mipBlitInfo = {
-        .srcImage = texture.m_Image,
-        .srcImageLayout = vk::ImageLayout::eTransferSrcOptimal,
-        .dstImage = texture.m_Image,
-        .dstImageLayout = vk::ImageLayout::eTransferDstOptimal,
-        .regionCount = 1,
-        .pRegions = &blitRegion,
-        .filter = vk::Filter::eLinear,
-    };
-
 #pragma endregion
 
     m_CommandBuffer.pipelineBarrier2(&imageStartDependency);
     m_CommandBuffer.copyBufferToImage2(&stagingCopyInfo);
-    m_CommandBuffer.pipelineBarrier2(&interMipDependency);
+    m_CommandBuffer.pipelineBarrier2(&postStagingDependency);
 
-    auto calcNextMip = [](i32 prev) { return eastl::max(prev / 2, 1); };
+    GenerateMipMaps(m_CommandBuffer, &texture, vk::ImageLayout::eTransferSrcOptimal,
+                    vk::ImageLayout::eShaderReadOnlyOptimal);
 
-    // Mip Mapping
-
-    i32 prevMipWidth = Cast<i32>(texture.m_Extent.width);
-    i32 prevMipHeight = Cast<i32>(texture.m_Extent.height);
-
-    u32 maxPrevMip = texture.GetMipLevels() - 1;
-    for (u32 prevMipLevel = 0; prevMipLevel < maxPrevMip; ++prevMipLevel)
-    {
-        i32 currentMipWidth = calcNextMip(prevMipWidth);
-        i32 currentMipHeight = calcNextMip(prevMipHeight);
-        u32 currentMipLevel = prevMipLevel + 1;
-
-        blitRegion.srcSubresource.mipLevel = prevMipLevel;
-        blitRegion.srcOffsets = std::array{
-            vk::Offset3D{0, 0, 0},
-            vk::Offset3D{prevMipWidth, prevMipHeight, 1},
-        };
-        blitRegion.dstSubresource.mipLevel = currentMipLevel;
-        blitRegion.dstOffsets = std::array{
-            vk::Offset3D{0, 0, 0},
-            vk::Offset3D{currentMipWidth, currentMipHeight, 1},
-        };
-
-        nextMipBarrier.subresourceRange.baseMipLevel = currentMipLevel;
-
-        m_CommandBuffer.blitImage2(&mipBlitInfo);
-        m_CommandBuffer.pipelineBarrier2(&interMipDependency);
-
-        prevMipHeight = currentMipHeight;
-        prevMipWidth = currentMipWidth;
-    }
-
-    m_CommandBuffer.pipelineBarrier2(&imageReadyDependency);
-
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     m_CommandBuffer.endDebugUtilsLabelEXT();
 #endif
 
-    return m_ResourceManager->Commit(&texture);
+    return m_ResourceManager->CommitTexture(&texture);
 }
 
 Model
@@ -425,7 +513,7 @@ AssetLoader::LoadModelToGpu(cstr path, cstr name)
 
     AbortIfFailed(m_CommandBuffer.begin(&OneTimeCmdBeginInfo));
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     StackString<128> loadActionName = "Load: ";
     loadActionName += name ? name : path;
     vk::DebugUtilsLabelEXT debugLabel = {
@@ -808,7 +896,7 @@ AssetLoader::LoadModelToGpu(cstr path, cstr name)
 
 #pragma endregion
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     m_CommandBuffer.endDebugUtilsLabelEXT();
 #endif
     AbortIfFailed(m_CommandBuffer.end());

samples/03_model_render/asset_loader.h

@@ -118,3 +118,8 @@ struct AssetLoader
 
     DISALLOW_COPY_AND_ASSIGN(AssetLoader);
 };
+
+void GenerateMipMaps(vk::CommandBuffer commandBuffer, Texture *texture, vk::ImageLayout initialLayout,
+                     vk::ImageLayout finalLayout,
+                     vk::PipelineStageFlags2 prevStage = vk::PipelineStageFlagBits2::eAllCommands,
+                     vk::PipelineStageFlags2 finalStage = vk::PipelineStageFlagBits2::eAllCommands);

samples/03_model_render/ibl_helpers.cpp

@@ -5,36 +5,86 @@
 
 #include "ibl_helpers.h"
 
+#include "EASTL/fixed_vector.h"
+#include "EASTL/tuple.h"
+#include "asset_loader.h"
 #include "device.h"
 #include "gpu_resource_manager.h"
 #include "helpers.h"
 #include "image.h"
-#include "asset_loader.h"
 #include "pipeline_utils.h"
 
-constexpr cstr EQUIRECT_TO_CUBE_SHADER_FILE = "shader/eqrectToCube.cs.hlsl.spv";
+constexpr cstr EQUIRECT_TO_CUBE_SHADER_FILE = "shader/eqrect_to_cube.cs.hlsl.spv";
+constexpr cstr DIFFUSE_IRRADIANCE_SHADER_FILE = "shader/diffuse_irradiance.cs.hlsl.spv";
+constexpr cstr PREFILTER_SHADER_FILE = "shader/prefilter.cs.hlsl.spv";
+constexpr cstr BRDF_LUT_SHADER_FILE = "shader/brdf_lut.cs.hlsl.spv";
 
-TextureHandle
+void
+Environment::Destroy(GpuResourceManager *resourceManager)
+{
+    resourceManager->Release(Take(m_Skybox));
+    resourceManager->Release(Take(m_Diffuse));
+    resourceManager->Release(Take(m_Prefilter));
+    resourceManager->Release(Take(m_BrdfLut));
+}
+
+Environment
 CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32 cubeSide, TextureHandle hdrEnv,
                      const cstr name)
 {
     GpuResourceManager *resMan = assetLoader->m_ResourceManager;
     const Device *pDevice = resMan->m_Device;
 
-    TextureCube cubeMap;
-    cubeMap.Init(pDevice, cubeSide, vk::Format::eR16G16B16A16Sfloat, false, name ? name : "Env CubeMap");
+    vk::SamplerCreateInfo brdfLutSamplerCreateInfo = resMan->m_DefaultSamplerCreateInfo;
+    brdfLutSamplerCreateInfo.addressModeU = vk::SamplerAddressMode::eClampToEdge;
+    brdfLutSamplerCreateInfo.addressModeV = vk::SamplerAddressMode::eClampToEdge;
+    brdfLutSamplerCreateInfo.addressModeW = vk::SamplerAddressMode::eClampToEdge;
 
-    StorageTexture stagingTexture;
-    stagingTexture.Init(pDevice, {cubeSide * 3, cubeSide * 2}, vk::Format::eR16G16B16A16Sfloat, false, "EnvStaging");
-    auto envStagingHandle = resMan->Commit(&stagingTexture);
+    StorageTextureCube skybox;
+    StorageTextureCube diffuseIrradiance;
+    StorageTextureCube prefilterCube;
+    StorageTexture brdfLut;
+    SamplerHandle brdfLutSampler;
 
-    vk::ImageSubresourceRange stagingSubresRange = {
-        .aspectMask = vk::ImageAspectFlagBits::eColor,
-        .baseMipLevel = 0,
-        .levelCount = 1,
-        .baseArrayLayer = 0,
-        .layerCount = 1,
-    };
+    skybox.Init(pDevice, cubeSide, vk::Format::eR16G16B16A16Sfloat, true, true, "Skybox");
+    TextureHandle skyboxHandle = resMan->CommitTexture(&skybox);
+    StorageTextureHandle skyboxStorageHandle = resMan->CommitStorageTexture(&skybox);
+
+    diffuseIrradiance.Init(pDevice, 64, vk::Format::eR16G16B16A16Sfloat, true, false, "Diffuse Irradiance");
+    TextureHandle diffuseIrradianceHandle = resMan->CommitTexture(&diffuseIrradiance);
+    StorageTextureHandle diffuseIrradianceStorageHandle = resMan->CommitStorageTexture(&diffuseIrradiance);
+
+    prefilterCube.Init(pDevice, cubeSide, vk::Format::eR16G16B16A16Sfloat, true, true, "Prefilter");
+    TextureHandle prefilterHandle = resMan->CommitTexture(&prefilterCube); // This stores the original view for us.
+    constexpr u32 prefilterMipCountMax = 6;
+    eastl::array<StorageTextureHandle, prefilterMipCountMax> prefilterStorageHandles;
+    // All non-owning copies.
+    for (u32 mipLevel = 0; auto &tex : prefilterStorageHandles)
+    {
+        vk::ImageViewCreateInfo imageViewCreateInfo = {
+            .image = prefilterCube.m_Image,
+            .viewType = vk::ImageViewType::eCube,
+            .format = vk::Format::eR16G16B16A16Sfloat,
+            .components = vk::ComponentMapping{},
+            .subresourceRange =
+                {
+                    .aspectMask = vk::ImageAspectFlagBits::eColor,
+                    .baseMipLevel = mipLevel++,
+                    .levelCount = 1,
+                    .baseArrayLayer = 0,
+                    .layerCount = 6,
+                },
+        };
+        AbortIfFailed(pDevice->m_Device.createImageView(&imageViewCreateInfo, nullptr, &prefilterCube.m_View));
+        tex = resMan->CommitStorageTexture(&prefilterCube);
+    }
+
+    brdfLut.Init(pDevice, {512, 512}, vk::Format::eR16G16Sfloat, true, "BRDF LUT");
+    brdfLutSampler = resMan->CreateSampler(&brdfLutSamplerCreateInfo);
+    TextureHandle brdfLutHandle = resMan->CommitTexture(&brdfLut, brdfLutSampler);
+    StorageTextureHandle brdfLutStorageHandle = resMan->CommitStorageTexture(&brdfLut);
+
+#pragma region Dependencies and Copies
     vk::ImageSubresourceRange cubeSubresRange = {
         .aspectMask = vk::ImageAspectFlagBits::eColor,
         .baseMipLevel = 0,
@@ -42,8 +92,15 @@ CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32
         .baseArrayLayer = 0,
         .layerCount = 6,
     };
+    vk::ImageSubresourceRange lutSubresRange = {
+        .aspectMask = vk::ImageAspectFlagBits::eColor,
+        .baseMipLevel = 0,
+        .levelCount = 1,
+        .baseArrayLayer = 0,
+        .layerCount = 1,
+    };
 
-    vk::ImageMemoryBarrier2 readyToWriteBarrier = {
+    vk::ImageMemoryBarrier2 readyToWriteBarrierTemplate = {
         .srcStageMask = vk::PipelineStageFlagBits2::eTopOfPipe,
         .srcAccessMask = vk::AccessFlagBits2::eNone,
         .dstStageMask = vk::PipelineStageFlagBits2::eComputeShader,
@@ -52,117 +109,94 @@ CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32
         .newLayout = vk::ImageLayout::eGeneral,
         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
         .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .image = stagingTexture.m_Image,
-        .subresourceRange = stagingSubresRange,
-    };
-    vk::DependencyInfo readyToWriteDependency = {
-        .imageMemoryBarrierCount = 1,
-        .pImageMemoryBarriers = &readyToWriteBarrier,
-    };
-    vk::ImageMemoryBarrier2 stagingReadyForTransfer = {
-        .srcStageMask = vk::PipelineStageFlagBits2::eComputeShader,
-        .srcAccessMask = vk::AccessFlagBits2::eShaderStorageWrite,
-        .dstStageMask = vk::PipelineStageFlagBits2::eAllTransfer,
-        .dstAccessMask = vk::AccessFlagBits2::eTransferRead,
-        .oldLayout = vk::ImageLayout::eGeneral,
-        .newLayout = vk::ImageLayout::eTransferSrcOptimal,
-        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .image = stagingTexture.m_Image,
-        .subresourceRange = stagingSubresRange,
-    };
-    vk::ImageMemoryBarrier2 cubeReadyForTransfer = {
-        .srcStageMask = vk::PipelineStageFlagBits2::eTopOfPipe,
-        .srcAccessMask = vk::AccessFlagBits2::eNone,
-        .dstStageMask = vk::PipelineStageFlagBits2::eAllTransfer,
-        .dstAccessMask = vk::AccessFlagBits2::eTransferWrite,
-        .oldLayout = vk::ImageLayout::eUndefined,
-        .newLayout = vk::ImageLayout::eTransferDstOptimal,
-        .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .image = cubeMap.m_Image,
         .subresourceRange = cubeSubresRange,
     };
-    eastl::array transferBarriers = {stagingReadyForTransfer, cubeReadyForTransfer};
-    vk::DependencyInfo preTransferDependency = {
-        .imageMemoryBarrierCount = transferBarriers.size(),
-        .pImageMemoryBarriers = transferBarriers.data(),
+    eastl::fixed_vector<vk::ImageMemoryBarrier2, 4> readyToWriteBarriers(4, readyToWriteBarrierTemplate);
+    readyToWriteBarriers[0].image = skybox.m_Image;
+    readyToWriteBarriers[1].image = diffuseIrradiance.m_Image;
+    readyToWriteBarriers[2].image = prefilterCube.m_Image;
+    readyToWriteBarriers[3].image = brdfLut.m_Image;
+    readyToWriteBarriers[3].subresourceRange = lutSubresRange;
+
+    vk::DependencyInfo readyToWriteDependency = {
+        .imageMemoryBarrierCount = Cast<u32>(readyToWriteBarriers.size()),
+        .pImageMemoryBarriers = readyToWriteBarriers.data(),
     };
-    vk::ImageMemoryBarrier2 cubemapToRead = {
-        .srcStageMask = vk::PipelineStageFlagBits2::eAllTransfer,
-        .srcAccessMask = vk::AccessFlagBits2::eTransferWrite,
-        .dstStageMask = vk::PipelineStageFlagBits2::eFragmentShader,
-        .dstAccessMask = vk::AccessFlagBits2::eShaderSampledRead,
-        .oldLayout = vk::ImageLayout::eTransferDstOptimal,
+
+    vk::ImageMemoryBarrier2 readyToSampleBarrierTemplate = {
+        .srcStageMask = vk::PipelineStageFlagBits2::eComputeShader,
+        .srcAccessMask = vk::AccessFlagBits2::eShaderStorageWrite | vk::AccessFlagBits2::eShaderStorageRead,
+        .dstStageMask = vk::PipelineStageFlagBits2::eBottomOfPipe,
+        .oldLayout = vk::ImageLayout::eGeneral,
         .newLayout = vk::ImageLayout::eShaderReadOnlyOptimal,
         .srcQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
         .dstQueueFamilyIndex = VK_QUEUE_FAMILY_IGNORED,
-        .image = cubeMap.m_Image,
         .subresourceRange = cubeSubresRange,
     };
-    vk::DependencyInfo cubemapToReadDependency = {
+    auto skyboxToSampleBarrier = readyToSampleBarrierTemplate;
+    skyboxToSampleBarrier.image = skybox.m_Image;
+
+    auto diffIrrToSampleBarrier = readyToSampleBarrierTemplate;
+    diffIrrToSampleBarrier.image = diffuseIrradiance.m_Image;
+
+    auto prefilterToSampleBarrier = readyToSampleBarrierTemplate;
+    prefilterToSampleBarrier.image = prefilterCube.m_Image;
+
+    auto brdfToSampleBarrier = readyToSampleBarrierTemplate;
+    prefilterToSampleBarrier.image = brdfLut.m_Image;
+    prefilterToSampleBarrier.subresourceRange = lutSubresRange;
+
+    vk::DependencyInfo skyboxToSampleDependency = {
         .imageMemoryBarrierCount = 1,
-        .pImageMemoryBarriers = &cubemapToRead,
+        .pImageMemoryBarriers = &skyboxToSampleBarrier,
+    };
+    vk::DependencyInfo diffIrrToSampleDependency = {
+        .imageMemoryBarrierCount = 1,
+        .pImageMemoryBarriers = &diffIrrToSampleBarrier,
+    };
+    vk::DependencyInfo prefilterToSampleDependency = {
+        .imageMemoryBarrierCount = 1,
+        .pImageMemoryBarriers = &prefilterToSampleBarrier,
+    };
+    vk::DependencyInfo brdfToSampleDependency = {
+        .imageMemoryBarrierCount = 1,
+        .pImageMemoryBarriers = &brdfToSampleBarrier,
     };
 
-    eastl::array<vk::ImageCopy2, 6> imageCopies;
-    for (i32 i = 0; i < 6; ++i)
-    {
-        imageCopies[i] = vk::ImageCopy2{
-            .srcSubresource =
-                {
-                    .aspectMask = vk::ImageAspectFlagBits::eColor,
-                    .mipLevel = 0,
-                    .baseArrayLayer = 0,
-                    .layerCount = 1,
-                },
-            .srcOffset =
-                {
-                    (i % 3) * Cast<i32>(cubeSide),
-                    (i / 3) * Cast<i32>(cubeSide),
-                    0,
-                },
-            .dstSubresource =
-                {
-                    .aspectMask = vk::ImageAspectFlagBits::eColor,
-                    .mipLevel = 0,
-                    .baseArrayLayer = Cast<u32>(i),
-                    .layerCount = 1,
-                },
-            .dstOffset = {0, 0, 0},
-            .extent = {cubeSide, cubeSide, 1},
-        };
-    }
+#pragma endregion
 
-    vk::CopyImageInfo2 copyFlatToCube = {
-        .srcImage = stagingTexture.m_Image,
-        .srcImageLayout = vk::ImageLayout::eTransferSrcOptimal,
-        .dstImage = cubeMap.m_Image,
-        .dstImageLayout = vk::ImageLayout::eTransferDstOptimal,
-        .regionCount = imageCopies.size(),
-        .pRegions = imageCopies.data(),
-    };
-
-    struct WorkloadPushConstants
+    struct SkyboxPushConstants
     {
         TextureHandle m_HdrEnvHandle;
         StorageTextureHandle m_OutputTexture;
         u32 m_CubeSide;
     };
+    struct DiffuseIrradiancePushConstants
+    {
+        TextureHandle m_SkyboxHandle;
+        StorageTextureHandle m_OutputTexture;
+        u32 m_CubeSide;
+    };
+    struct PrefilterPushConstants
+    {
+        TextureHandle m_SkyboxHandle;
+        StorageTextureHandle m_OutputTexture;
+        u32 m_CubeSide;
+        f32 m_Roughness;
+        u32 m_EnvSide;
+    };
+    struct BrdfLutPushConstants
+    {
+        StorageTextureHandle m_OutputTexture;
+    };
 
 #pragma region Pipeline Creation etc
-    const auto shaderModule = CreateShader(pDevice, EQUIRECT_TO_CUBE_SHADER_FILE);
-
-    const vk::PipelineShaderStageCreateInfo shaderStageCreateInfo = {
-        .stage = vk::ShaderStageFlagBits::eCompute,
-        .module = shaderModule,
-        .pName = "main",
-    };
 
     vk::PushConstantRange pcr = {
         .stageFlags = vk::ShaderStageFlagBits::eCompute,
         .offset = 0,
-        .size = sizeof(WorkloadPushConstants),
+        .size = eastl::max(eastl::max(sizeof(SkyboxPushConstants), sizeof(BrdfLutPushConstants)),
+                           eastl::max(sizeof(DiffuseIrradiancePushConstants), sizeof(PrefilterPushConstants))),
     };
 
     vk::PipelineLayout pipelineLayout;
@@ -174,21 +208,83 @@ CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32
     };
     AbortIfFailed(pDevice->m_Device.createPipelineLayout(&layoutCreateInfo, nullptr, &pipelineLayout));
 
-    const vk::ComputePipelineCreateInfo computePipelineCreateInfo = {
-        .stage = shaderStageCreateInfo,
-        .layout = pipelineLayout,
+    const auto eqRectToCubeShader = CreateShader(pDevice, EQUIRECT_TO_CUBE_SHADER_FILE);
+    const auto diffuseRadianceShader = CreateShader(pDevice, DIFFUSE_IRRADIANCE_SHADER_FILE);
+    const auto prefilterShader = CreateShader(pDevice, PREFILTER_SHADER_FILE);
+    const auto brdfLutShader = CreateShader(pDevice, BRDF_LUT_SHADER_FILE);
+    eastl::array computePipelineCreateInfo = {
+        vk::ComputePipelineCreateInfo{
+            .stage =
+                {
+                    .stage = vk::ShaderStageFlagBits::eCompute,
+                    .module = eqRectToCubeShader,
+                    .pName = "main",
+                },
+            .layout = pipelineLayout,
+        },
+        vk::ComputePipelineCreateInfo{
+            .stage =
+                {
+                    .stage = vk::ShaderStageFlagBits::eCompute,
+                    .module = diffuseRadianceShader,
+                    .pName = "main",
+                },
+            .layout = pipelineLayout,
+        },
+        vk::ComputePipelineCreateInfo{
+            .stage =
+                {
+                    .stage = vk::ShaderStageFlagBits::eCompute,
+                    .module = prefilterShader,
+                    .pName = "main",
+                },
+            .layout = pipelineLayout,
+        },
+        vk::ComputePipelineCreateInfo{
+            .stage =
+                {
+                    .stage = vk::ShaderStageFlagBits::eCompute,
+                    .module = brdfLutShader,
+                    .pName = "main",
+                },
+            .layout = pipelineLayout,
+        },
     };
 
-    vk::Pipeline pipeline;
-    AbortIfFailed(pDevice->m_Device.createComputePipelines(pDevice->m_PipelineCache, 1, &computePipelineCreateInfo,
-                                                           nullptr, &pipeline));
+    eastl::array<vk::Pipeline, computePipelineCreateInfo.size()> pipelines;
+    AbortIfFailed(pDevice->m_Device.createComputePipelines(pDevice->m_PipelineCache, computePipelineCreateInfo.size(),
+                                                           computePipelineCreateInfo.data(), nullptr,
+                                                           pipelines.data()));
 
-    pDevice->m_Device.destroy(shaderModule, nullptr);
+    vk::Pipeline eqRectToCubePipeline = pipelines[0];
+    vk::Pipeline diffuseIrradiancePipeline = pipelines[1];
+    vk::Pipeline prefilterPipeline = pipelines[2];
+    vk::Pipeline brdfLutPipeline = pipelines[3];
+
+    for (auto &createInfos : computePipelineCreateInfo)
+    {
+        pDevice->m_Device.destroy(createInfos.stage.module, nullptr);
+    }
 
 #pragma endregion
 
-    WorkloadPushConstants pushConstants = {
-        .m_HdrEnvHandle = hdrEnv, .m_OutputTexture = envStagingHandle, .m_CubeSide = cubeSide};
+    SkyboxPushConstants skyboxPushConstant = {
+        .m_HdrEnvHandle = hdrEnv,
+        .m_OutputTexture = skyboxStorageHandle,
+        .m_CubeSide = cubeSide,
+    };
+    DiffuseIrradiancePushConstants diffuseIrradiancePushConstants = {
+        .m_SkyboxHandle = skyboxHandle,
+        .m_OutputTexture = diffuseIrradianceStorageHandle,
+        .m_CubeSide = diffuseIrradiance.m_Extent.width,
+    };
+    PrefilterPushConstants prefilterPushConstants = {
+        .m_SkyboxHandle = skyboxHandle,
+        .m_EnvSide = cubeSide,
+    };
+    BrdfLutPushConstants brdfLutPushConstants = {
+        .m_OutputTexture = brdfLutStorageHandle,
+    };
 
     resMan->Update();
 
@@ -196,7 +292,7 @@ CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32
     constexpr vk::CommandBufferBeginInfo beginInfo = {.flags = vk::CommandBufferUsageFlagBits::eOneTimeSubmit};
     AbortIfFailed(cmd.begin(&beginInfo));
 
-#if !defined(NDEBUG)
+#if !defined(ASTER_NDEBUG)
     StackString<128> labelName = "Eqrect -> Cubemap: ";
     labelName += name ? name : "<unknown env>";
     vk::DebugUtilsLabelEXT label = {
@@ -209,17 +305,50 @@ CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32
     cmd.pipelineBarrier2(&readyToWriteDependency);
 
     cmd.bindDescriptorSets(vk::PipelineBindPoint::eCompute, pipelineLayout, 0, 1, &resMan->m_DescriptorSet, 0, nullptr);
-    cmd.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline);
-    cmd.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof pushConstants, &pushConstants);
-    cmd.dispatch(cubeSide / 16, cubeSide / 16, 6);
+    cmd.bindPipeline(vk::PipelineBindPoint::eCompute, eqRectToCubePipeline);
+    cmd.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof skyboxPushConstant,
+                      &skyboxPushConstant);
+    assert(skybox.m_Extent.width % 16 == 0 && skybox.m_Extent.height % 16 == 0);
+    cmd.dispatch(skybox.m_Extent.width / 16, skybox.m_Extent.height / 16, 6);
 
-    cmd.pipelineBarrier2(&preTransferDependency);
+    GenerateMipMaps(cmd, &skybox, vk::ImageLayout::eGeneral, vk::ImageLayout::eGeneral,
+                    vk::PipelineStageFlagBits2::eComputeShader, vk::PipelineStageFlagBits2::eComputeShader);
 
-    cmd.copyImage2(&copyFlatToCube);
+    cmd.bindPipeline(vk::PipelineBindPoint::eCompute, diffuseIrradiancePipeline);
+    cmd.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof skyboxPushConstant,
+                      &diffuseIrradiancePushConstants);
+    assert(diffuseIrradiance.m_Extent.width % 16 == 0 && diffuseIrradiance.m_Extent.height % 16 == 0);
+    cmd.dispatch(diffuseIrradiance.m_Extent.width / 16, diffuseIrradiance.m_Extent.width / 16, 6);
 
-    cmd.pipelineBarrier2(&cubemapToReadDependency);
+    cmd.pipelineBarrier2(&diffIrrToSampleDependency);
 
-#if !defined(NDEBUG)
+    cmd.bindPipeline(vk::PipelineBindPoint::eCompute, prefilterPipeline);
+    u32 mipSize = prefilterCube.m_Extent.width;
+    assert(mipSize % 16 == 0);
+    for (u32 mipCount = 0; auto &tex : prefilterStorageHandles)
+    {
+        prefilterPushConstants.m_OutputTexture = tex;
+        prefilterPushConstants.m_CubeSide = mipSize;
+        prefilterPushConstants.m_Roughness = Cast<f32>(mipCount) / Cast<f32>(prefilterMipCountMax);
+        cmd.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof prefilterPushConstants,
+                          &prefilterPushConstants);
+        u32 groupCount = eastl::max(mipSize / 16u, 1u);
+        cmd.dispatch(groupCount, groupCount, 6);
+
+        ++mipCount;
+        mipSize = mipSize >> 1;
+    }
+
+    cmd.pipelineBarrier2(&skyboxToSampleDependency);
+    cmd.pipelineBarrier2(&prefilterToSampleDependency);
+
+    cmd.bindPipeline(vk::PipelineBindPoint::eCompute, brdfLutPipeline);
+    cmd.pushConstants(pipelineLayout, vk::ShaderStageFlagBits::eCompute, 0, sizeof brdfLutPushConstants,
+                      &brdfLutPushConstants);
+    assert(brdfLut.m_Extent.width % 16 == 0 && brdfLut.m_Extent.height % 16 == 0);
+    cmd.dispatch(brdfLut.m_Extent.width / 16, brdfLut.m_Extent.height / 16, 1);
+
+#if !defined(ASTER_NDEBUG)
     cmd.endDebugUtilsLabelEXT();
 #endif
 
@@ -241,9 +370,26 @@ CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, const u32
 
     AbortIfFailed(pDevice->m_Device.resetCommandPool(assetLoader->m_CommandPool, {}));
 
-    resMan->Release(envStagingHandle);
-    pDevice->m_Device.destroy(pipeline, nullptr);
+    skybox = {};
+    resMan->Release(skyboxStorageHandle);
+    resMan->Release(diffuseIrradianceStorageHandle);
+    resMan->Release(brdfLutStorageHandle);
+    for (auto &texHandles : prefilterStorageHandles)
+    {
+        StorageTextureCube st;
+        resMan->Release(&st, texHandles);
+        pDevice->m_Device.destroy(st.m_View, nullptr);
+    }
+    for (auto &pipeline : pipelines)
+    {
+        pDevice->m_Device.destroy(pipeline, nullptr);
+    }
     pDevice->m_Device.destroy(pipelineLayout, nullptr);
 
-    return resMan->Commit(&cubeMap);
+    return {
+        .m_Skybox = skyboxHandle,
+        .m_Diffuse = diffuseIrradianceHandle,
+        .m_Prefilter = prefilterHandle,
+        .m_BrdfLut = brdfLutHandle,
+    };
 }

samples/03_model_render/ibl_helpers.h

@@ -8,8 +8,23 @@
 #include "global.h"
 #include "gpu_resource_manager.h"
 
+#include <EASTL/tuple.h>
+
+struct Pipeline;
 struct Texture;
 struct TextureCube;
 struct AssetLoader;
 
-TextureHandle CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, u32 cubeSide, TextureHandle hdrEnv, cstr name = nullptr);
+struct Environment
+{
+    TextureHandle m_Skybox;
+    TextureHandle m_Diffuse;
+    TextureHandle m_Prefilter;
+    TextureHandle m_BrdfLut;
+
+    void Destroy(GpuResourceManager *resourceManager);
+};
+
+Environment
+CreateCubeFromHdrEnv(AssetLoader *assetLoader, vk::Queue computeQueue, u32 cubeSide, TextureHandle hdrEnv,
+                                   cstr name = nullptr);

samples/03_model_render/model_render.cpp

@@ -18,10 +18,10 @@
 #include "helpers.h"
 #include "light_manager.h"
 
+#include "asset_loader.h"
 #include "gpu_resource_manager.h"
 #include "gui.h"
 #include "ibl_helpers.h"
-#include "asset_loader.h"
 #include "pipeline_utils.h"
 
 #include <EASTL/array.h>
@@ -176,13 +176,13 @@ main(int, char **)
     LightManager lightManager = LightManager{&resourceManager};
 
     Model model = assetLoader.LoadModelToGpu(MODEL_FILE);
-    Texture environment;
-    assetLoader.LoadHdrImage(&environment, BACKDROP_FILE);
-    auto envHandle = resourceManager.Commit(&environment);
+    Texture environmentHdri;
+    assetLoader.LoadHdrImage(&environmentHdri, BACKDROP_FILE);
+    auto envHdriHandle = resourceManager.CommitTexture(&environmentHdri);
 
-    TextureHandle texCube = CreateCubeFromHdrEnv(&assetLoader, graphicsQueue, 1024, envHandle, "Cube Env");
+    auto environment = CreateCubeFromHdrEnv(&assetLoader, graphicsQueue, 512, envHdriHandle, "Cube Env");
 
-    resourceManager.Release(envHandle);
+    resourceManager.Release(envHdriHandle);
 
     vk::Format attachmentFormat = vk::Format::eR8G8B8A8Srgb;
 
@@ -371,6 +371,11 @@ main(int, char **)
               queueAllocation.m_Family, graphicsQueue);
     bool rotating = false;
     bool lockToScreen = true;
+    bool showDiffuse = false;
+    bool useDiffuse = true;
+    bool showPrefilter = false;
+    bool useSpecular = true;
+
     i32 height = Cast<i32>(internalResolution.height);
     f32 camPitch = glm::degrees(cameraController.m_Pitch);
     f32 camYaw = glm::degrees(cameraController.m_Yaw);
@@ -445,6 +450,13 @@ main(int, char **)
         {
             cameraController.SetPosition(camPosition);
         }
+        gui::Separator();
+        gui::Text("IBL");
+        gui::Checkbox("Show DiffIrr", &showDiffuse);
+        gui::Checkbox("Use DiffIrr", &useDiffuse);
+        gui::Checkbox("Show Prefilter", &showPrefilter);
+        gui::Checkbox("Use Specular", &useSpecular);
+        gui::Separator();
         gui::Checkbox("Rotate", &rotating);
         gui::PopItemWidth();
         if (gui::Button("Exit"))
@@ -546,8 +558,32 @@ main(int, char **)
         cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, pcbOffset, sizeof model.m_Handles,
                           &model.m_Handles);
         pcbOffset += sizeof model.m_Handles;
-        cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, pcbOffset, sizeof texCube, &texCube);
-        pcbOffset += sizeof texCube;
+
+        Environment thisFrameEnvBuffers = environment;
+        if (showDiffuse)
+        {
+            thisFrameEnvBuffers.m_Skybox = environment.m_Diffuse;
+        }
+        else if (showPrefilter)
+        {
+            thisFrameEnvBuffers.m_Skybox = environment.m_Prefilter;
+        }
+
+        if (!useDiffuse)
+        {
+            thisFrameEnvBuffers.m_Diffuse = {};
+        }
+        if (!useSpecular)
+        {
+            thisFrameEnvBuffers.m_BrdfLut = {};
+            thisFrameEnvBuffers.m_Prefilter = {};
+        }
+
+        assert(pcbOffset == 16);
+        cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, pcbOffset, sizeof thisFrameEnvBuffers,
+                          &thisFrameEnvBuffers);
+        pcbOffset += sizeof thisFrameEnvBuffers;
+
         cmd.pushConstants(pipeline.m_Layout, vk::ShaderStageFlagBits::eAll, pcbOffset, sizeof lightManager.m_MetaInfo,
                           &lightManager.m_MetaInfo);
         pcbOffset += sizeof lightManager.m_MetaInfo;
@@ -625,7 +661,7 @@ main(int, char **)
 
     AbortIfFailed(device.m_Device.waitIdle());
 
-    resourceManager.Release(texCube);
+    environment.Destroy(&resourceManager);
 
     pipelineCacheData = device.DumpPipelineCache();
     ERROR_IF(!WriteFileBytes(PIPELINE_CACHE_FILE, pipelineCacheData), "Pipeline Cache incorrectly written");

samples/03_model_render/pipeline_utils.cpp

@@ -57,7 +57,7 @@ CreatePipeline(const Device *device, vk::Format attachmentFormat, const GpuResou
     vk::PushConstantRange pushConstantRange = {
         .stageFlags = vk::ShaderStageFlagBits::eAll,
         .offset = 0,
-        .size = 40,
+        .size = 52,
     };
 
     vk::PipelineLayoutCreateInfo pipelineLayoutCreateInfo = {
@@ -206,7 +206,7 @@ CreateBackgroundPipeline(const Device *device, vk::Format attachmentFormat, cons
     vk::PushConstantRange pushConstantRange = {
         .stageFlags = vk::ShaderStageFlagBits::eAll,
         .offset = 0,
-        .size = 40,
+        .size = 52,
     };
 
     vk::PipelineLayoutCreateInfo pipelineLayoutCreateInfo = {

samples/03_model_render/shader/bindless_structs.hlsli

@@ -48,7 +48,9 @@ struct DirectionalLight
 
 #define INVALID_HANDLE 0xFFFFFFFF
 
+static const float HALF_PI = 1.57079633f;
 static const float PI = 3.14159265f;
+static const float TAU = 6.28318530f;
 
 [[vk::binding(0, 0)]] StructuredBuffer<float4> VertexBuffer[];
 [[vk::binding(0, 0)]] StructuredBuffer<VertexData> VertexDataBuffer[];
@@ -58,7 +60,10 @@ static const float PI = 3.14159265f;
 [[vk::binding(0, 0)]] StructuredBuffer<DirectionalLight> DirectionalLightBuffer[];
 
 [[vk::binding(1, 0)]] Texture2D<float4> Textures[];
+[[vk::binding(1, 0)]] Texture2D<float2> TexturesRG[];
 [[vk::binding(1, 0)]] TextureCube<float4> TextureCubes[];
 [[vk::binding(1, 0)]] SamplerState ImmutableSamplers[];
 
 [[vk::binding(2, 0)]] RWTexture2D<float4> StorageTextures[];
+[[vk::binding(2, 0)]] RWTexture2D<float2> StorageTexturesRG[];
+[[vk::binding(2, 0)]] RWTexture2DArray<float4> StorageTextureArrays[];

samples/03_model_render/shader/brdf_lut.cs.hlsl

@@ -0,0 +1,82 @@
+#include "ibl_common.hlsli"
+
+struct Block
+{
+    uint OutputTextureHandle;
+};
+
+[[vk::push_constant]]
+Block Pcb;
+
+float GeometrySchlickGGX(float NdotV, float Roughness)
+{
+    float R = Roughness;
+    // (Rough + 1)^2 / 8 for Punctual Lights
+    // Rough^2 / 2 for IBL
+    float K = (R * R) / 2.0;
+
+    float Numerator = NdotV;
+    float Denominator = NdotV * (1.0f - K) + K;
+	
+    return Numerator / Denominator;
+}
+
+float GeometrySmith(float NdotV, float NdotL, float Roughness)
+{
+    float GGX1 = GeometrySchlickGGX(NdotV, Roughness);
+    float GGX2 = GeometrySchlickGGX(NdotL, Roughness);
+	
+    return GGX1 * GGX2;
+}
+          
+float2 IntegrateBRDF(float NdotV, float Roughness)
+{
+    float3 ViewDir;
+    ViewDir.x = sqrt(1.0f - NdotV * NdotV);
+    ViewDir.y = 0.0f;
+    ViewDir.z = NdotV;
+
+    float A = 0.0f;
+    float B = 0.0f;
+
+    float3 Normal = float3(0.0f, 0.0f, 1.0f);
+
+    const uint SAMPLE_COUNT = 1024u;
+    
+    for (uint i = 0u; i < SAMPLE_COUNT; ++i)
+    {
+        float2 Xi = Hammersley(i, SAMPLE_COUNT);
+        float3 Halfway = ImportanceSampleGGX(Xi, Normal, Roughness);
+        float3 LightDir = normalize(2.0f * dot(ViewDir, Halfway) * Halfway - ViewDir);
+
+        float NdotL = max(LightDir.z, 0.0f);
+        float NdotH = max(Halfway.z, 0.0f);
+        float VdotH = max(dot(ViewDir, Halfway), 0.0f);
+
+        if (NdotL > 0.0f)
+        {
+            float G = GeometrySmith(NdotV, NdotL, Roughness);
+            float G_Vis = (G * VdotH) / max((NdotH * NdotV), 0.0001f);
+            float Fc = pow(1.0f - VdotH, 5.0f);
+
+            A += (1.0f - Fc) * G_Vis;
+            B += Fc * G_Vis;
+        }
+    }
+    A /= float(SAMPLE_COUNT);
+    B /= float(SAMPLE_COUNT);
+    
+    return float2(A, B);
+}
+
+[numthreads(16, 16, 1)]
+void main(uint3 GlobalInvocationID : SV_DispatchThreadID)
+{
+    float Width, Height;
+    StorageTexturesRG[Pcb.OutputTextureHandle].GetDimensions(Width, Height);
+    
+    float2 UV = GlobalInvocationID.xy / float2(Width, Height);
+    
+    float2 IntegratedBRDF = IntegrateBRDF(UV.x, UV.y);
+    StorageTexturesRG[Pcb.OutputTextureHandle][GlobalInvocationID.xy] = IntegratedBRDF;
+}

samples/03_model_render/shader/diffuse_irradiance.cs.hlsl

@@ -0,0 +1,52 @@
+#include "ibl_common.hlsli"
+
+struct Block
+{
+    uint SkyboxHandle;
+    uint OutputTextureHandle;
+    int CubeSide;
+};
+
+[[vk::push_constant]]
+Block pcb;
+
+/*
+| Axis | Layer | Up |
+|:----:|:-----:|:--:|
+|  +x  |   0   | -y |
+|  -x  |   1   | -y |
+|  +y  |   2   | +z |
+|  -y  |   3   | -z |
+|  -z  |   4   | -y |
+|  +z  |   5   | -y |
+*/
+
+[numthreads(16, 16, 1)]
+void main(uint3 GlobalInvocationID : SV_DispatchThreadID)
+{
+    float3 Forward, Up, Right;
+    
+    Forward = GetCubeDir(GlobalInvocationID, pcb.CubeSide);
+    Up = abs(Forward.y) < 1.0f ? float3(0.0f, 1.0f, 0.0f) : float3(1.0f, 0.0f, 0.0f); // 0.01f offset to 
+    Right = normalize(cross(Up, Forward));
+    Up = normalize(cross(Forward, Right));
+    
+    float3 Irradiance = 0.0f.xxx;
+    float3 IrrDirr = 0.0f.xxx;
+    float SampleStep = 0.005f;
+    float SampleCount = 0.0f;
+  
+    for (float Azimuth = 0.0f; Azimuth < TAU; Azimuth += SampleStep)
+    {
+        for (float Zenith = 0.0f; Zenith < HALF_PI; Zenith += SampleStep)
+        {
+            float3 DirectionTanSpace = float3(sin(Zenith) * cos(Azimuth), sin(Zenith) * sin(Azimuth), cos(Zenith));
+            float3 DirectionWorld = DirectionTanSpace.x * Right + DirectionTanSpace.y * Up + DirectionTanSpace.z * Forward;
+            
+            Irradiance += TextureCubes[pcb.SkyboxHandle].SampleLevel(ImmutableSamplers[pcb.SkyboxHandle], DirectionWorld, 0).xyz * (cos(Zenith) * sin(Zenith));
+            SampleCount++;
+        }
+    }
+            
+    StorageTextureArrays[pcb.OutputTextureHandle][GlobalInvocationID.xyz] = PI * float4(Irradiance * (1.0f / SampleCount), 1.0f);
+}

samples/03_model_render/shader/eqrectToCube.cs.hlsl

@@ -1,58 +0,0 @@
-#include "bindless_structs.hlsli"
-
-struct Block
-{
-    uint HdrEnvHandle;
-    uint OutputTextureHandle;
-    int CubeSize;
-};
-
-[[vk::push_constant]]
-Block pcb;
-
-float2 SampleSphericalMap(float3 v)
-{
-    const float2 invAtan = float2(0.1591f, 0.3183f);
-    float2 uv = float2(atan2(v.x, v.z), asin(v.y));
-    uv *= invAtan;
-    uv += 0.5;
-    return uv;
-}
-
-/*
-| Axis | Layer | Up |
-|:----:|:-----:|:--:|
-|  +x  |   0   | -y |
-|  -x  |   1   | -y |
-|  +y  |   2   | +z |
-|  -y  |   3   | -z |
-|  -z  |   4   | -y |
-|  +z  |   5   | -y |
-*/
-
-[numthreads(16, 16, 1)]
-void main(uint3 GlobalInvocationID : SV_DispatchThreadID)
-{
-    float3 LocalDir;
-    float HalfSide = pcb.CubeSize / 2;
-    float AxisSign = 2.0f * (0.5f - (GlobalInvocationID.z % 2));
-    
-    if (GlobalInvocationID.z < 2)
-    {
-        LocalDir = float3(AxisSign * HalfSide, GlobalInvocationID.y - HalfSide, (GlobalInvocationID.x - HalfSide) * AxisSign);
-    }
-    else if (GlobalInvocationID.z < 4)
-    {
-        LocalDir = float3(GlobalInvocationID.x - HalfSide, -AxisSign * HalfSide, -AxisSign * (GlobalInvocationID.y - HalfSide));
-    }
-    else
-    {
-        LocalDir = float3((GlobalInvocationID.x - HalfSide) * AxisSign, GlobalInvocationID.y - HalfSide, -AxisSign * HalfSide);
-    }
-    
-    uint TileX = GlobalInvocationID.z % 3;
-    uint TileY = GlobalInvocationID.z / 3;
-    
-    float2 UV = SampleSphericalMap(normalize(LocalDir));
-    StorageTextures[pcb.OutputTextureHandle][GlobalInvocationID.xy + uint2(TileX, TileY) * pcb.CubeSize] = Textures[pcb.HdrEnvHandle].SampleLevel(ImmutableSamplers[pcb.HdrEnvHandle], UV, 0);
-}

samples/03_model_render/shader/eqrect_to_cube.cs.hlsl

@@ -0,0 +1,29 @@
+#include "ibl_common.hlsli"
+
+struct Block
+{
+    uint HdrEnvHandle;
+    uint OutputTextureHandle;
+    int CubeSide;
+};
+
+[[vk::push_constant]]
+Block pcb;
+
+float2 SampleSphericalMap(float3 v)
+{
+    const float2 InvTan = float2(0.1591f, 0.3183f);     // (1/2PI, 1/PI)
+    float2 UV = float2(atan2(-v.x, v.z), asin(-v.y));   // (-PI, -PI/2) to (PI, PI/2)
+    UV *= InvTan;                                       // (-1/2, -1/2) to (1/2, 1/2)
+    UV += 0.5f.xx;                                      // (0, 0) to (1, 1)
+    return UV;
+}
+
+[numthreads(16, 16, 1)]
+void main(uint3 GlobalInvocationID : SV_DispatchThreadID)
+{
+    float3 LocalDir = GetCubeDir(GlobalInvocationID, pcb.CubeSide);
+    
+    float2 UV = SampleSphericalMap(LocalDir);
+    StorageTextureArrays[pcb.OutputTextureHandle][GlobalInvocationID.xyz] = Textures[pcb.HdrEnvHandle].SampleLevel(ImmutableSamplers[pcb.HdrEnvHandle], UV, 0);
+}

samples/03_model_render/shader/graphics_structs.hlsli

@@ -15,10 +15,16 @@ struct Block
     uint VertexDataHandle;
     uint MaterialBufferHandle;
     uint NodeBufferHandle;
+    
     uint EnvCubeHandle;
+    uint DiffuseIrradianceHandle;
+    uint PrefilterHandle;
+    uint BrdfLutHandle;
+    
     uint LightHandle;
     uint PointLightIndexer;
     uint DirectionalLightIndexer;
+    
     int MaterialIdx;
     uint NodeIdx;
 };

samples/03_model_render/shader/ibl_common.hlsli

@@ -0,0 +1,77 @@
+#include "bindless_structs.hlsli"
+
+/*
+ The Goal is simply to convert from a (0,0,0) to (CubeSide, CubeSide, FaceCount) Invocation ID space to a
+ (-1,-1,-1) to (1, 1, 1) space.
+
+| Axis | Layer | Up |
+|:----:|:-----:|:--:|
+|  +x  |   0   | -y |
+|  -x  |   1   | -y |
+|  +y  |   2   | +z |
+|  -y  |   3   | -z |
+|  -z  |   4   | -y |
+|  +z  |   5   | -y |
+*/
+float3 GetCubeDir(uint3 GlobalInvocationID, float SideLength)
+{
+    float2 FaceUV = float2(GlobalInvocationID.xy) / SideLength; // (0, SideLength) -> (0, 1)
+    FaceUV = 2.0f * FaceUV - 1.0f; // (0, 1) -> (-1, 1)
+    
+    switch (GlobalInvocationID.z)
+    {
+        case 0:
+            return normalize(float3(1.0f, -FaceUV.y, -FaceUV.x)); // Face +X; x = 1, y = -v, z = -u
+        case 1:
+            return normalize(float3(-1.0f, -FaceUV.y, FaceUV.x)); // Face -X; x = -1, y = -v, z = u
+        case 2:
+            return normalize(float3(FaceUV.x, 1.0f, FaceUV.y)); // Face +Y; x = u, y = 1, z = v
+        case 3:
+            return normalize(float3(FaceUV.x, -1.0f, -FaceUV.y)); // Face -Y; x=u, y=-1, z=-v
+        case 4:
+            return normalize(float3(FaceUV.x, -FaceUV.y, 1.0f)); // Face +Z; x=u,y=-v, z=1
+        case 5:
+            return normalize(float3(-FaceUV.x, -FaceUV.y, -1.0f)); // Face -Z; x=u,y=-v, z=-1
+        default:
+            // Never reach here.
+            return 0.0f.xxx;
+    }
+}
+
+float RadicalInverse_VdC(uint Bits)
+{
+    Bits = (Bits << 16u) | (Bits >> 16u);
+    Bits = ((Bits & 0x55555555u) << 1u) | ((Bits & 0xAAAAAAAAu) >> 1u);
+    Bits = ((Bits & 0x33333333u) << 2u) | ((Bits & 0xCCCCCCCCu) >> 2u);
+    Bits = ((Bits & 0x0F0F0F0Fu) << 4u) | ((Bits & 0xF0F0F0F0u) >> 4u);
+    Bits = ((Bits & 0x00FF00FFu) << 8u) | ((Bits & 0xFF00FF00u) >> 8u);
+    return float(Bits) * 2.3283064365386963e-10; // / 0x100000000
+}
+
+float2 Hammersley(uint SampleIndex, uint SampleCount)
+{
+    return float2(float(SampleIndex) / float(SampleCount), RadicalInverse_VdC(SampleIndex));
+}
+
+float3 ImportanceSampleGGX(float2 Xi, float3 Normal, float Roughness)
+{
+    float A = Roughness * Roughness;
+	
+    float Phi = 2.0f * PI * Xi.x;
+    float CosTheta = sqrt((1.0f - Xi.y) / (1.0f + (A * A - 1.0f) * Xi.y));
+    float SinTheta = sqrt(1.0f - CosTheta * CosTheta);
+	
+	// from spherical coordinates to cartesian coordinates
+    float3 H;
+    H.x = cos(Phi) * SinTheta;
+    H.y = sin(Phi) * SinTheta;
+    H.z = CosTheta;
+	
+	// from tangent-space vector to world-space sample vector
+    float3 Up = abs(Normal.z) < 0.999f ? float3(0.0f, 0.0f, 1.0f) : float3(1.0f, 0.0f, 0.0f);
+    float3 Tangent = normalize(cross(Up, Normal));
+    float3 Bitangent = cross(Normal, Tangent);
+	
+    float3 SampleVec = Tangent * H.x + Bitangent * H.y + Normal * H.z;
+    return normalize(SampleVec);
+}

samples/03_model_render/shader/model.ps.hlsl

@@ -10,7 +10,7 @@ struct FS_Input
 
 struct FS_Output
 {
-    float4 ColorTarget  : SV_Target0;
+    float4 ColorTarget : SV_Target0;
 };
 
 float4 GetAlbedo(float2 UV, float4 InColor)
@@ -21,6 +21,13 @@ float4 GetAlbedo(float2 UV, float4 InColor)
     return AlbedoFactor * InColor * (AlbedoTexId != INVALID_HANDLE ? Textures[AlbedoTexId].Sample(ImmutableSamplers[AlbedoTexId], UV) : 1.0f.xxxx);
 }
 
+float GetOcclusion(float2 UV)
+{
+    uint OcclusionTexId = MaterialsBuffer[PushConstant.MaterialBufferHandle][PushConstant.MaterialIdx].OcclusionTex;
+    
+    return OcclusionTexId != INVALID_HANDLE ? Textures[OcclusionTexId].Sample(ImmutableSamplers[OcclusionTexId], UV).r : 1.0f;
+}
+
 float3 GetEmissive(float2 UV)
 {
     float3 EmissionFactor = (float3) MaterialsBuffer[PushConstant.MaterialBufferHandle][PushConstant.MaterialIdx].EmissionFactor;
@@ -61,6 +68,35 @@ float2 GetMetalRough(float2 UV)
     return MetalRoughFactors * (MetalRoughTexId != INVALID_HANDLE ? Textures[MetalRoughTexId].Sample(ImmutableSamplers[MetalRoughTexId], UV).bg : 1.0f.xx); // Metal is B, Rough is G.
 }
 
+float3 SampleIrradiance(float3 Direction)
+{
+    if (PushConstant.DiffuseIrradianceHandle != INVALID_HANDLE)
+    {
+        return TextureCubes[PushConstant.DiffuseIrradianceHandle].Sample(ImmutableSamplers[PushConstant.DiffuseIrradianceHandle], Direction).rgb;
+    }
+    return 0.04f.xxx;
+}
+
+float3 SamplePrefiltered(float3 Direction, float Roughness)
+{
+    const float MAX_MIP_LEVEL = 5.0f;
+    float Mip = MAX_MIP_LEVEL * Roughness;
+    if (PushConstant.PrefilterHandle != INVALID_HANDLE)
+    {
+        return TextureCubes[PushConstant.PrefilterHandle].SampleLevel(ImmutableSamplers[PushConstant.PrefilterHandle], Direction, Mip).rgb;
+    }
+    return 0.0f.xxx;
+}
+
+float2 SampleBrdfLut(float NdotV, float Roughness)
+{
+    if (PushConstant.BrdfLutHandle != INVALID_HANDLE)
+    {
+        return TexturesRG[PushConstant.BrdfLutHandle].Sample(ImmutableSamplers[PushConstant.BrdfLutHandle], float2(NdotV, Roughness));
+    }
+    return 0.0f.xx;
+}
+
 float TrowbridgeReitzGGX(float3 Normal, float3 Halfway, float Roughness)
 {
     float Coeff = Roughness * Roughness;
@@ -100,6 +136,12 @@ float3 FresnelSchlick(float cosine, float3 F_0)
     return F_0 + (1.0f - F_0) * pow(clamp(1.0f - cosine, 0.0f, 1.0f), 5.0f); // Clamp to avoid artifacts.
 }
 
+// Sebastian Lagarde
+float3 FresnelSchlickRoughness(float cosine, float3 F_0, float Roughness)
+{
+    return F_0 + (max((1.0f - Roughness).xxx, F_0) - F_0) * pow(clamp(1.0f - cosine, 0.0f, 1.0f), 5.0f); // Clamp to avoid artifacts.
+}
+
 float3 GetPBRContrib(float3 Albedo, float3 LightColor, float3 ViewDir, float3 Normal, float Metallic, float Roughness, float3 F_0, float3 LightDir, float LightDistance)
 {
     float Attenuation = 1.0f / (LightDistance * LightDistance); // TODO: Controlled Attenuation
@@ -113,18 +155,18 @@ float3 GetPBRContrib(float3 Albedo, float3 LightColor, float3 ViewDir, float3 No
     
     float NormalDistribution = TrowbridgeReitzGGX(Normal, Halfway, Roughness);
     float Geometry = GeometrySmith(NdotV, NdotL, Roughness);
-    float3 Fresnel = FresnelSchlick(CosineFactor, F_0);
+    float3 Fresnel = FresnelSchlickRoughness(CosineFactor, F_0, Roughness);
         
     float3 Numerator = (NormalDistribution * Geometry) * Fresnel;
     float Denominator = 4.0f * NdotV * NdotL;
-    float3 Specular = Numerator / max(Denominator, 0.00001);
+    float3 Specular = Numerator / (Denominator + 0.00001f);
         
     float3 K_Specular = Fresnel;
     float3 K_Diffuse = 1.0f.xxx - K_Specular;
         
     K_Diffuse *= 1.0f - Metallic;
         
-    return NdotL * Radiance * ((K_Diffuse * Albedo / PI) + Specular);
+    return NdotL * Radiance * (K_Diffuse * Albedo / PI + Specular);
 }
 
 float3 GetPointLightInfluence(float3 Albedo, float2 MetalRough, float3 Position, float3 Normal)
@@ -144,7 +186,6 @@ float3 GetPointLightInfluence(float3 Albedo, float2 MetalRough, float3 Position,
     // TODO: Cite    
     float3 F_0 = 0.04f.xxx;
     F_0 = lerp(F_0, Albedo, Metallic);
-    float3 LightAmp = float3(23.47f, 21.31f, 20.79f); // TODO: Get rid
     
     float3 Contrib = 0.0f;
     for (uint i = 0; i < Count; ++i)
@@ -167,7 +208,7 @@ float3 GetPointLightInfluence(float3 Albedo, float2 MetalRough, float3 Position,
         float G = (Light.Color & 0x00FF0000) >> 16;
         float B = (Light.Color & 0x0000FF00) >> 8;
         
-        float3 LightColor = LightAmp * Light.Intensity * float3(R, G, B) * 0.00392156862f; // 0.00392156862 = 1/255
+        float3 LightColor = Light.Intensity * float3(R, G, B) * 0.00392156862f; // 0.00392156862 = 1/255
  
         Contrib += GetPBRContrib(Albedo, LightColor, ViewDir, Normal, Metallic, Roughness, F_0, LightDir, LightDistance);
     }
@@ -175,7 +216,6 @@ float3 GetPointLightInfluence(float3 Albedo, float2 MetalRough, float3 Position,
     return Contrib;
 }
 
-
 float3 GetDirectionalLightInfluence(float3 Albedo, float2 MetalRough, float3 Position, float3 Normal)
 {
     if (PushConstant.LightHandle == INVALID_HANDLE) 
@@ -192,7 +232,6 @@ float3 GetDirectionalLightInfluence(float3 Albedo, float2 MetalRough, float3 Pos
     // TODO: Cite    
     float3 F_0 = 0.04f.xxx;
     F_0 = lerp(F_0, Albedo, Metallic);
-    float3 LightAmp = float3(23.47f, 21.31f, 20.79f); // TODO: Get rid
     
     float3 Contrib = 0.0f;
     for (uint i = 0; i < Count; ++i)
@@ -209,7 +248,7 @@ float3 GetDirectionalLightInfluence(float3 Albedo, float2 MetalRough, float3 Pos
         float G = (Light.Color & 0x00FF0000) >> 16;
         float B = (Light.Color & 0x0000FF00) >> 8;
         
-        float3 LightColor = LightAmp * Light.Intensity * float3(R, G, B) * 0.00392156862f; // 0.00392156862 = 1/255
+        float3 LightColor = Light.Intensity * float3(R, G, B) * 0.00392156862f; // 0.00392156862 = 1/255
  
         Contrib += GetPBRContrib(Albedo, LightColor, ViewDir, Normal, Metallic, Roughness, F_0, LightDir, 1.0f);
     }
@@ -217,12 +256,37 @@ float3 GetDirectionalLightInfluence(float3 Albedo, float2 MetalRough, float3 Pos
     return Contrib;
 }
 
+float3 GetAmbientInfluence(float3 Albedo, float2 MetalRough, float3 Position, float3 Normal, float Occlusion)
+{
+    float3 ViewDir = normalize(Camera.Position.xyz - Position);
+    float CosineFactor = max(dot(Normal, ViewDir), 0.0f);       // Normal instead of Halfway since there's no halfway in ambient.
+    
+    float Metal = MetalRough.r;
+    float Roughness = MetalRough.g;
+    
+    float3 F_0 = 0.04f.xxx;
+    F_0 = lerp(F_0, Albedo, MetalRough.r);
+    float3 K_Specular = FresnelSchlickRoughness(CosineFactor, F_0, Roughness);
+    float3 K_Diffuse = 1.0f.xxx - K_Specular;
+        
+    K_Diffuse *= 1.0f - Metal; // Metals don't have diffuse/refractions.
+    
+    float3 ReflectionDir = reflect(-ViewDir, Normal);
+    
+    float NdotV = max(dot(Normal, ViewDir), 0.0f);
+    float3 PrefilteredColor = SamplePrefiltered(ReflectionDir, Roughness).rgb;
+    float2 EnvBRDF = SampleBrdfLut(NdotV, Roughness);
+    float3 Specular = PrefilteredColor * (K_Specular * EnvBRDF.x + EnvBRDF.y);
+    
+    float3 DiffuseIrradiance = Albedo * SampleIrradiance(Normal);
+    
+    return (K_Diffuse * DiffuseIrradiance + Specular) * Occlusion;
+}
+
 FS_Output main(FS_Input StageInput)
 {
     FS_Output Output;
     
-    float3 Ambient = float3(0.02f, 0.02f, 0.02f);
-    
     // TODO: This should be invalid on the CPU side.
     if (PushConstant.MaterialIdx < 0)
     {
@@ -238,7 +302,12 @@ FS_Output main(FS_Input StageInput)
     float Alpha = AlbedoAlpha.a;
     float2 MetalRough = GetMetalRough(StageInput.InUV0);
     float3 Emission = GetEmissive(StageInput.InUV0);
-    float3 Color = GetDirectionalLightInfluence(Albedo, MetalRough, Position, Normal) + GetPointLightInfluence(Albedo, MetalRough, Position, Normal);
+    float Occlusion = GetOcclusion(StageInput.InUV0);
+    
+    float3 DirectionalLightLum = GetDirectionalLightInfluence(Albedo, MetalRough, Position, Normal);
+    float3 PointLighLum = GetPointLightInfluence(Albedo, MetalRough, Position, Normal);
+    float3 AmbientLum = GetAmbientInfluence(Albedo, MetalRough, Position, Normal, Occlusion);
+    float3 Color = DirectionalLightLum + PointLighLum + AmbientLum;
     
     Output.ColorTarget = float4(Uncharted2Tonemap(Color + Emission), Alpha);
     return Output;

samples/03_model_render/shader/prefilter.cs.hlsl

@@ -0,0 +1,68 @@
+#include "ibl_common.hlsli"
+
+struct Block
+{
+    uint SkyboxHandle;
+    uint OutputTextureHandle;
+    int CubeSide;
+    float Roughness;
+    uint EnvRes;
+};
+
+[[vk::push_constant]]
+Block Pcb;
+
+float TrowbridgeReitzGGX(float NdotH, float Roughness)
+{
+    float Coeff = Roughness * Roughness;
+    float Alpha2 = Coeff * Coeff;
+    float NdotH2 = NdotH * NdotH;
+	
+    float Numerator = Alpha2;
+    float Denominator = NdotH2 * (Alpha2 - 1.0f) + 1.0f;
+    Denominator = PI * Denominator * Denominator;
+	
+    return Numerator / Denominator;
+}
+
+float GetSampleMipLevel(float NdotH, float HdotV, float SampleCount)
+{
+    float D = TrowbridgeReitzGGX(NdotH, Pcb.Roughness);
+    float Pdf = (D * NdotH / (4.0f * HdotV)) + 0.0001f;
+
+    float SurfAreaTexel = 4.0f * PI / (6.0f * Pcb.EnvRes * Pcb.EnvRes);
+    float SurfAreaSample = 1.0f / (SampleCount * Pdf + 0.0001f);
+
+    return Pcb.Roughness == 0.0f ? 0.0f : 0.5f * log2(SurfAreaSample / SurfAreaTexel);
+}
+
+[numthreads(16, 16, 1)]
+void main(uint3 GlobalInvocationID : SV_DispatchThreadID)
+{   
+    float3 Normal = GetCubeDir(GlobalInvocationID, Pcb.CubeSide);
+    float3 ViewDir = Normal;
+
+    const uint SAMPLE_COUNT = 2048u;
+    float TotalWeight = 0.0f;
+    float3 PrefilterColor = 0.0f.xxx;
+    for (uint i = 0u; i < SAMPLE_COUNT; ++i)
+    {
+        float2 Xi = Hammersley(i, SAMPLE_COUNT);
+        float3 Halfway = ImportanceSampleGGX(Xi, Normal, Pcb.Roughness);
+        float3 LightDir = normalize(2.0 * dot(ViewDir, Halfway) * Halfway - ViewDir);
+        
+        float NdotH = max(dot(Normal, Halfway), 0.0f);
+    
+        float MipLevel = GetSampleMipLevel(NdotH, NdotH /* N = V :: NdotH = HdotV */, SAMPLE_COUNT);
+        
+        float NdotL = max(dot(Normal, LightDir), 0.0);
+        if (NdotL > 0.0)
+        {
+            PrefilterColor += TextureCubes[Pcb.SkyboxHandle].SampleLevel(ImmutableSamplers[Pcb.SkyboxHandle], LightDir, MipLevel).rgb * NdotL;
+            TotalWeight += NdotL;
+        }
+    }
+    PrefilterColor = PrefilterColor / TotalWeight;
+
+    StorageTextureArrays[Pcb.OutputTextureHandle][GlobalInvocationID] = float4(PrefilterColor, 1.0f);
+}